PageSourceSearch

https://agentos.sh/blog/

html agentos.sh collected 2026-10-05 06:27:54 UTC 188,448 bytes, 84 lines download raw bytes

1<!DOCTYPE html><html lang="en"><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=5"/><link rel="preload" href="/_next/static/media/e4af272ccee01ff0-s.p.woff2" as="font" crossorigin="" type="font/woff2"/><link rel="preload" as="image" href="/img/blog/paracosm/sim-poster.jpg"/><link rel="preload" as="image" href="/img/blog/og/agentos-memory-sota-longmemeval.png"/><link rel="stylesheet" href="/_next/static/css/34251478a1fd3bc2.css" data-precedence="next"/><link rel="stylesheet" href="/_next/static/css/62f1ca5a4f071d84.css" data-precedence="next"/><link rel="stylesheet" href="/_next/static/css/96b04247e785c9c6.css" data-precedence="next"/><link rel="preload" as="script" fetchPriority="low" href="/_next/static/chunks/webpack-bf5df1ec8c5aa8d1.js"/>
1<script src="/_next/static/chunks/lib-b0bd9fc7c9825fdb.js" async=""></script>
1<script src="/_next/static/chunks/main-app-ab1b128c4e42b4e3.js" async=""></script>
1<script src="/_next/static/chunks/app/%5Blocale%5D/loading-547537401f418075.js" async=""></script>
1<script src="/_next/static/chunks/app/layout-991763ab09c81468.js" async=""></script>
1<script src="/_next/static/chunks/app/loading-7b32c3a4f4ce7f0e.js" async=""></script>
1<script src="/_next/static/chunks/app/%5Blocale%5D/blog/page-a30ab686764a69d2.js" async=""></script>
1<script src="/_next/static/chunks/7586-203f1916d23f2407.js" async=""></script>
1<script src="/_next/static/chunks/9079-00ef10f8b11f4d94.js" async=""></script>
1<script src="/_next/static/chunks/1230-cd3b090f825a7d79.js" async=""></script>
1<script src="/_next/static/chunks/4400-4ff185dce52c1c17.js" async=""></script>
1<script src="/_next/static/chunks/7531-64596f3089411931.js" async=""></script>
1<script src="/_next/static/chunks/app/%5Blocale%5D/page-45ab204874fc5d47.js" async=""></script>
1<script src="/_next/static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js" async=""></script>
1<title>AgentOS Blog | TypeScript AI Agent Framework Tutorials &amp; Benchmarks</title><meta name="description" content="Technical articles, framework comparisons, tutorials, and engineering insights from the AgentOS team. Learn how to build production AI agents with TypeScript."/><link rel="author" href="https://frame.dev"/><meta name="author" content="Frame.dev"/><meta name="keywords" content="AgentOS,agentos,AI agent framework,TypeScript AI agent framework,AI agent SDK TypeScript,open source AI agent framework,build AI agents,build AI agents TypeScript,AI agent engineering platform,autonomous AI agents,production-ready AI agents,multi-agent orchestration,multi-agent system,agentic AI framework,agentic workflows,agent simulation framework,AI agent simulation,cognitive AI agent,AI agent memory system,adaptive intelligence,emergent behaviors,AI guardrails,prompt injection defense,cognitive memory,multimodal RAG,RAG pipeline TypeScript,voice AI agents,AI agent orchestration patterns,graph orchestration,AI workflow automation,self-hosted AI agents,agent runtime,runtime tool forging,self-improving agents,type-safe AI agents,LangGraph alternative,CrewAI alternative,AutoGen alternative,Mastra alternative,VoltAgent alternative,Vercel AI SDK alternative,OpenAI Agents SDK alternative"/><meta name="creator" content="Frame.dev"/><meta name="publisher" content="Frame.dev"/><meta name="robots" content="index, follow"/><meta name="googlebot" content="index, follow, max-video-preview:-1, max-image-preview:large, max-snippet:-1"/><link rel="canonical" href="https://agentos.sh/blog/"/><link rel="alternate" hrefLang="en" href="https://agentos.sh/blog/"/><link rel="alternate" hrefLang="es" href="https://agentos.sh/es/blog/"/><link rel="alternate" hrefLang="fr" href="https://agentos.sh/fr/blog/"/><link rel="alternate" hrefLang="de" href="https://agentos.sh/de/blog/"/><link rel="alternate" hrefLang="pt" href="https://agentos.sh/pt/blog/"/><link rel="alternate" hrefLang="ja" href="https://agentos.sh/ja/blog/"/><link rel="alternate" hrefLang="ko" href="https://agentos.sh/ko/blog/"/><link rel="alternate" hrefLang="zh" href="https://agentos.sh/zh/blog/"/><link rel="alternate" hrefLang="x-default" href="https://agentos.sh/blog/"/><meta property="og:title" content="AgentOS Blog | TypeScript AI Agent Framework Tutorials &amp; Benchmarks"/><meta property="og:description" content="Technical articles, framework comparisons, tutorials, and engineering insights from the AgentOS team."/><meta property="og:url" content="https://agentos.sh/blog/"/><meta property="og:site_name" content="AgentOS"/><meta property="og:image" content="https://agentos.sh/og-image-v2.png"/><meta property="og:type" content="website"/><meta name="twitter:card" content="summary_large_image"/><meta name="twitter:title" content="AgentOS Blog | TypeScript AI Agent Framework Tutorials &amp; Benchmarks"/><meta name="twitter:description" content="Technical articles, framework comparisons, and tutorials for building AI agents with TypeScript."/><meta name="twitter:image" content="https://agentos.sh/og-image-v2.png"/><link rel="shortcut icon" href="/favicon.svg"/><link rel="icon" href="/favicon.svg" type="image/svg+xml"/><link rel="icon" href="/icon-128.png" sizes="128x128" type="image/png"/><link rel="icon" href="/icon-256.png" sizes="256x256" type="image/png"/><link rel="icon" href="/icon-512.png" sizes="512x512" type="image/png"/><link rel="apple-touch-icon" href="/icon-256.png" sizes="256x256" type="image/png"/><meta name="next-size-adjust"/><link rel="icon" href="/favicon.ico"/><link rel="icon" type="image/svg+xml" href="/favicon.svg"/><link rel="preconnect" href="https://api.github.com" crossorigin="anonymous"/><link rel="preconnect" href="https://img.shields.io" crossorigin="anonymous"/><link rel="dns-prefetch" href="https://api.npmjs.org"/><link rel="dns-prefetch" href="https://static.cloudflareinsights.com"/><link rel="dns-prefetch" href="https://www.googletagmanager.com"/><link rel="dns-prefetch" href="https://www.google-analytics.com"/>
1<script>
2              (function() {
3                try {
4                  var THEME_VERSION = '2026-04-29-twilight-default';
5                  var modeKey = 'agentos-mode';
6                  var themeKey = 'agentos-theme';
7                  var versionKey = 'agentos-theme-version';
8
9                  var ver = localStorage.getItem(versionKey);
10                  if (ver !== THEME_VERSION) {
11                    localStorage.setItem(versionKey, THEME_VERSION);
12                    localStorage.removeItem(themeKey);
13                  }
14
15                  var mode = localStorage.getItem(modeKey);
16                  if (!mode || mode === 'system') {
17                    mode = window.matchMedia && window.matchMedia('(prefers-color-scheme: dark)').matches ? 'dark' : 'light';
18                  }
19                  if (mode === 'dark') {
20                    document.documentElement.classList.add('dark');
21                    document.documentElement.style.colorScheme = 'dark';
22                  } else {
23                    document.documentElement.classList.remove('dark');
24                    document.documentElement.style.colorScheme = 'light';
25                  }
26
27                  var theme = localStorage.getItem(themeKey) || 'twilight-neo';
28                  document.documentElement.setAttribute('data-theme', theme);
29
30                  // BenchmarkBanner dismissal — sync-check before paint so
31                  // returning visitors don't see the banner flash in then
32                  // out (the React useState/useEffect path renders visible
33                  // → measures localStorage → hides, producing a ~70px CLS
34                  // shift Lighthouse attributed 0.16 to on mobile). Keep
35                  // the storage key in sync with the BenchmarkBanner
36                  // STORAGE_KEY constant.
37                  if (localStorage.getItem('agentos-banner-2026-04-29') === 'dismissed') {
38                    document.documentElement.setAttribute('data-banner-dismissed', 'true');
39                  }
40                } catch (e) { /* swallow: localStorage may be unavailable in some embeds */ }
41              })();
42            </script>
42<style>
43          * { box-sizing: border-box; }
44          html, body { width: 100%; margin: 0; padding: 0; }
45          html { height: 100%; overflow-x: clip; }
46          body { min-height: 100vh; overflow-x: clip; }
47
48          .skip-to-content { position: absolute; left: -9999px; z-index: 999; padding: 1rem 1.5rem; background: hsl(250, 95%, 55%); color: white; text-decoration: none; border-radius: 0.5rem; font-weight: 600; }
49          .skip-to-content:focus { left: 1rem; top: 1rem; outline: 2px solid hsl(280, 85%, 60%); outline-offset: 2px; }
50          .skeleton { position: relative; overflow: hidden; border-radius: 0.5rem; }
51          .skeleton::after { content: ""; position: absolute; inset: 0; background: linear-gradient(90deg, transparent 0%, rgba(200,200,200,0.3) 50%, transparent 100%); animation: skeleton-shimmer 2s ease-in-out infinite; }
52          @keyframes skeleton-shimmer { 0% { transform: translateX(-100%); } 100% { transform: translateX(100%); } }
53
54          .hero-critical { display: flex; flex-direction: column; justify-content: center; position: relative; min-height: 100vh; overflow: hidden; }
55          .hero-critical > * { position: relative; z-index: 1; }
56
57          /* Banner dismissal — paired with the bootstrap script that sets
58             data-banner-dismissed before paint. The banner's own
59             useState(false) default would otherwise render visible then
60             vanish on hydration, shifting the hero down by ~70px. */
61          html[data-banner-dismissed="true"] [aria-label="Latest benchmark result"] { display: none; }
62
63          @media (prefers-reduced-motion: reduce) {
64            *, *::before, *::after {
65              animation-duration: 0.01ms !important;
66              animation-iteration-count: 1 !important;
67              transition-duration: 0.01ms !important;
68              scroll-behavior: auto !important;
69            }
70          }
71        </style>
71<script src="/_next/static/chunks/polyfills-42372ed130431b0a.js" noModule=""></script>
71</head><body class="__variable_7748ca grainy min-h-screen antialiased transition-theme bg-background-primary text-text-primary"><!--$-->
71<script>!function(){try{var d=document.documentElement,c=d.classList;c.remove('light','dark');var e=localStorage.getItem('agentos-mode');if('system'===e||(!e&&true)){var t='(prefers-color-scheme: dark)',m=window.matchMedia(t);if(m.media!==t||m.matches){d.style.colorScheme = 'dark';c.add('dark')}else{d.style.colorScheme = 'light';c.add('light')}}else if(e){c.add(e|| '')}if(e==='light'||e==='dark')d.style.colorScheme=e}catch(e){}}()</script>
71<script type="application/ld+json">{"@context":"https://schema.org","@type":"SoftwareApplication","name":"AgentOS","alternateName":["AgentOS Framework","AgentOS Runtime","AgentOS by Frame"],"description":"Open-source TypeScript AI agent framework. Build autonomous agents with cognitive memory, runtime tool generation, multi-agent orchestration, multimodal RAG, voice pipelines, and 11 LLM providers.","url":"https://agentos.sh","applicationCategory":"DeveloperApplication","applicationSubCategory":"AI Agent Framework","operatingSystem":"Any","programmingLanguage":"TypeScript","keywords":"TypeScript AI agent framework, AI agent runtime, open-source agent framework, multi-agent orchestration, cognitive memory, agent SDK, autonomous agents, AgentOS","offers":{"@type":"Offer","price":"0","priceCurrency":"USD","description":"Free and open-source (Apache 2.0)"},"author":{"@type":"Organization","name":"Frame","url":"https://frame.dev"},"downloadUrl":"https://www.npmjs.com/package/@framers/agentos","installUrl":"https://www.npmjs.com/package/@framers/agentos","softwareVersion":"0.1.x","license":"https://opensource.org/licenses/Apache-2.0","image":"https://agentos.sh/og-image-v2.png","screenshot":"https://agentos.sh/og-image-v2.png","featureList":["TypeScript-first AI agent framework","Multi-agent orchestration with 6 collaboration strategies","Cognitive memory with Ebbinghaus decay and reconsolidation","Runtime tool generation: agents write their own functions mid-task","Specialist spawning: spawn_specialist mints new agents at runtime","Multimodal RAG (text, image, audio, video)","11 LLM providers (OpenAI, Anthropic, Google, Ollama, OpenRouter)","37 channel adapters (Telegram, WhatsApp, Discord, Slack)","Built-in safety: PII redaction, prompt injection defense, content moderation","HEXACO personality system","Voice pipeline (STT, TTS, VAD)","80+ curated skills and extensions","Self-hostable, production-ready, Apache 2.0"],"sameAs":["https://github.com/framerslab/agentos","https://www.npmjs.com/package/@framers/agentos","https://docs.agentos.sh"]}</script>
71<script type="application/ld+json">{"@context":"https://schema.org","@type":"Organization","name":"Frame","alternateName":["Frame.dev","Frame team"],"url":"https://frame.dev","logo":"https://agentos.sh/logo.png","email":"mailto:[email protected]","sameAs":["https://github.com/framerslab","https://twitter.com/framerslab","https://www.linkedin.com/company/framerslab","https://agentos.sh","https://vca.chat"]}</script>
71<script type="application/ld+json">{"@context":"https://schema.org","@type":"WebSite","name":"AgentOS","url":"https://agentos.sh","potentialAction":{"@type":"SearchAction","target":"https://docs.agentos.sh/search?q={search_term_string}","query-input":"required name=search_term_string"}}</script>
71<!--$!--><template data-dgst="BAILOUT_TO_CLIENT_SIDE_RENDERING"></template><!--/$--><!--$!--><template data-dgst="BAILOUT_TO_CLIENT_SIDE_RENDERING"></template><!--/$--><!--$!--><template data-dgst="BAILOUT_TO_CLIENT_SIDE_RENDERING"></template><!--/$--><a href="#main-content" class="skip-to-content">Skip to content</a><!--$--><header class="fixed top-0 z-50 w-full animate-fade-in" style="will-change:transform"><div class="absolute inset-0 backdrop-blur-xl border-b border-transparent shadow-[0_10px_35px_rgba(4,6,15,0.45)]" style="background-color:var(--color-background-primary);background-image:linear-gradient(135deg, color-mix(in oklab, var(--color-background-primary) 90%, transparent), color-mix(in oklab, var(--color-accent-primary) 16%, transparent) 55%, color-mix(in oklab, var(--color-accent-secondary) 12%, transparent))"></div><div class="pointer-events-none absolute bottom-0 left-0 right-0 h-[1px] opacity-70" style="background:linear-gradient(90deg, color-mix(in oklab, var(--color-accent-primary) 55%, transparent), color-mix(in oklab, var(--color-accent-secondary) 35%, transparent) 50%, color-mix(in oklab, var(--color-accent-primary) 55%, transparent));box-shadow:0 0 24px color-mix(in oklab, var(--color-accent-primary) 25%, transparent)"></div><div class="relative z-10 w-full"><div class="mx-auto flex w-full max-w-7xl items-center justify-between px-2 sm:px-5 lg:px-6 py-3"><a aria-label="AgentOS home" class="group flex items-center gap-2 transition-all hover:scale-[1.02] leading-none" href="/en/"><div class="relative overflow-visible"><div class="absolute inset-0 bg-accent-primary/20 blur-xl rounded-full animate-pulse-glow"></div><span class="theme-logo-wordmark text-gray-900 dark:text-white h-10 relative z-10 transform scale-[1.18]" aria-hidden="true"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 320 80" width="160" height="40" role="presentation"><defs><linearGradient id="agentos-gradient--Rist9ukq-" x1="0%" y1="0%" x2="100%" y2="100%"><stop offset="0%" stop-color="#6366F1"></stop><stop offset="50%" stop-color="#8B5CF6"></stop><stop offset="100%" stop-color="#EC4899"></stop></linearGradient></defs><g><circle cx="30" cy="40" r="6" fill="url(#agentos-gradient--Rist9ukq-)" opacity="1"></circle><circle cx="15" cy="25" r="4" fill="#6366F1" opacity="0.8"></circle><circle cx="45" cy="25" r="4" fill="#8B5CF6" opacity="0.8"></circle><circle cx="50" cy="45" r="4" fill="#EC4899" opacity="0.8"></circle><circle cx="30" cy="58" r="4" fill="#06B6D4" opacity="0.8"></circle><circle cx="10" cy="45" r="4" fill="#6366F1" opacity="0.8"></circle><line x1="30" y1="40" x2="15" y2="25" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="1.5" opacity="0.6"></line><line x1="30" y1="40" x2="45" y2="25" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="1.5" opacity="0.6"></line><line x1="30" y1="40" x2="50" y2="45" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="1.5" opacity="0.6"></line><line x1="30" y1="40" x2="30" y2="58" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="1.5" opacity="0.6"></line><line x1="30" y1="40" x2="10" y2="45" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="1.5" opacity="0.6"></line><line x1="15" y1="25" x2="45" y2="25" stroke="#6366F1" stroke-width="1" opacity="0.3"></line><line x1="10" y1="45" x2="50" y2="45" stroke="#8B5CF6" stroke-width="1" opacity="0.3"></line><circle cx="30" cy="40" r="12" fill="none" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="0.5" opacity="0.3"></circle><circle cx="30" cy="40" r="18" fill="none" stroke="url(#agentos-gradient--Rist9ukq-)" stroke-width="0.3" opacity="0.2"></circle></g><text x="70" y="45" font-family="var(--font-inter), -apple-system, BlinkMacSystemFont, &#x27;Segoe UI&#x27;, sans-serif" font-size="32" font-weight="600" fill="currentColor">Agent<tspan fill="url(#agentos-gradient--Rist9ukq-)">OS</tspan></text></svg></span><span class="sr-only">AgentOS</span></div></a><nav class="hidden items-center gap-5 lg:gap-7 text-sm font-medium lg:flex" aria-label="Main navigation"><a class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]" href="/en/features/"><span class="relative z-10 transition-all duration-300 ease-out">Features</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a><a class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]" href="/en/about/"><span class="relative z-10 transition-all duration-300 ease-out">About</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a><a href="https://paracosm.agentos.sh" class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]"><span class="relative z-10 transition-all duration-300 ease-out">Paracosm</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a><a href="https://docs.agentos.sh/" class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]"><span class="relative z-10 transition-all duration-300 ease-out">Docs</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a><a class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]" href="/en/blog/"><span class="relative z-10 transition-all duration-300 ease-out">Blog</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a><a class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]" href="/en/faq/"><span class="relative z-10 transition-all duration-300 ease-out">FAQ</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a><a class="nav-link group relative inline-block whitespace-nowrap py-2 px-1 text-[var(--color-text-primary)] font-semibold transition-all duration-300 ease-out hover:text-[var(--color-accent-primary)]" href="/en/contact/"><span class="relative z-10 transition-all duration-300 ease-out">Contact</span><span class="absolute bottom-0 left-0 h-[2px] w-0 bg-gradient-to-r from-[var(--color-accent-primary)] to-[var(--color-accent-secondary)] transition-all duration-500 ease-out group-hover:w-full"></span></a></nav><div class="flex items-center gap-1 sm:gap-2"><button type="button" class="hidden sm:inline-flex items-center justify-center p-2 rounded-xl text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/10 transition-all duration-300" aria-label="Search documentation" title="Search (/ or Ctrl+K)"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-search w-4 h-4"><circle cx="11" cy="11" r="8"></circle><path d="m21 21-4.3-4.3"></path></svg></button><a href="https://github.com/framerslab/agentos" class="relative hidden lg:inline-flex items-center gap-2 px-3 py-2 rounded-full border border-border-subtle text-[var(--color-text-primary)] hover:text-accent-primary transition-all duration-300 hover:-translate-y-0.5 group" target="_blank" rel="noopener noreferrer" aria-label="Open AgentOS on GitHub" title="Open AgentOS on GitHub"><span class="pointer-events-none absolute inset-0 rounded-full opacity-0 transition-opacity duration-300 group-hover:opacity-100" style="background:linear-gradient(90deg, color-mix(in oklab, var(--color-accent-primary) 18%, transparent), color-mix(in oklab, var(--color-accent-secondary) 14%, transparent))"></span><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-github w-4 h-4" aria-hidden="true"><path d="M15 22v-4a4.8 4.8 0 0 0-1-3.5c3 0 6-2 6-5.5.08-1.25-.27-2.48-1-3.5.28-1.15.28-2.35 0-3.5 0 0-1 0-3 1.5-2.64-.5-5.36-.5-8 0C6 2 5 2 5 2c-.3 1.15-.3 2.35 0 3.5A5.403 5.403 0 0 0 4 9c0 3.5 3 5.5 6 5.5-.39.49-.68 1.05-.85 1.65-.17.6-.22 1.23-.15 1.85v4"></path><path d="M9 18c-4.51 2-5-2-7-2"></path></svg><span class="font-semibold">GitHub</span></a><div class="hidden sm:flex items-center gap-1.5"><div class="relative"><button class="inline-flex items-center justify-center rounded-lg border border-slate-300 bg-white/80 p-2 text-sm text-slate-700 shadow-sm transition hover:bg-white hover:shadow-md dark:border-slate-600 dar
71k:bg-slate-800/80 dark:text-slate-200 dark:hover:bg-slate-800" aria-label="Select theme" aria-expanded="false"><svg viewBox="0 0 24 24" class="h-5 w-5" aria-hidden="true"><circle cx="12" cy="12" r="4" fill="#06b6d4" opacity="0.9"></circle><circle cx="12" cy="12" r="8" fill="none" stroke="#8b5cf6" stroke-width="1.5" opacity="0.5"></circle><line x1="12" y1="4" x2="12" y2="1" stroke="#06b6d4" stroke-width="1.5" opacity="0.7"></line><line x1="20" y1="12" x2="23" y2="12" stroke="#8b5cf6" stroke-width="1.5" opacity="0.7"></line><line x1="12" y1="20" x2="12" y2="23" stroke="#06b6d4" stroke-width="1.5" opacity="0.7"></line><line x1="4" y1="12" x2="1" y2="12" stroke="#8b5cf6" stroke-width="1.5" opacity="0.7"></line></svg></button></div><button class="flex h-9 w-9 items-center justify-center rounded-lg border border-slate-300 bg-white/80 text-slate-700 shadow-sm transition-theme dark:border-slate-600 dark:bg-slate-800/80 dark:text-slate-200" aria-label="Toggle theme"><div class="h-4 w-4 rounded-full bg-gradient-to-tr from-purple-300 to-pink-300"></div></button><div class="relative"><button type="button" class="inline-flex items-center gap-2 px-3 py-2 rounded-xl glass-morphism hover:bg-accent-primary/10 transition-all min-h-[44px] min-w-[44px]" aria-label="Select language" aria-expanded="false" aria-haspopup="true"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-globe w-4 h-4"><circle cx="12" cy="12" r="10"></circle><path d="M12 2a14.5 14.5 0 0 0 0 20 14.5 14.5 0 0 0 0-20"></path><path d="M2 12h20"></path></svg><span class="hidden sm:inline text-sm font-medium">English</span></button></div></div><div class="flex sm:hidden items-center"><button class="flex h-9 w-9 items-center justify-center rounded-lg border border-slate-300 bg-white/80 text-slate-700 shadow-sm transition-theme dark:border-slate-600 dark:bg-slate-800/80 dark:text-slate-200" aria-label="Toggle theme"><div class="h-4 w-4 rounded-full bg-gradient-to-tr from-purple-300 to-pink-300"></div></button></div><button type="button" aria-label="Open navigation menu" aria-expanded="false" aria-controls="mobile-menu" class="inline-flex items-center justify-center p-2 rounded-xl glass-morphism hover:bg-accent-primary/10 transition-all lg:hidden min-h-[44px] min-w-[44px]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-menu h-6 w-6"><line x1="4" x2="20" y1="12" y2="12"></line><line x1="4" x2="20" y1="6" y2="6"></line><line x1="4" x2="20" y1="18" y2="18"></line></svg></button></div></div></div><div class="lg:hidden transition-all duration-200 opacity-0 -translate-y-4 pointer-events-none" id="mobile-menu" style="display:none"><div class="mx-4 mt-2 rounded-2xl glass-morphism shadow-modern overflow-hidden"><nav class="flex flex-col" aria-label="Mobile navigation"><a class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0" href="/en/features/">Features</a><a class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0" href="/en/about/">About</a><a href="https://paracosm.agentos.sh" class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0">Paracosm</a><a href="https://docs.agentos.sh/" class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0">Docs</a><a class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0" href="/en/blog/">Blog</a><a class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0" href="/en/faq/">FAQ</a><a class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out block border-b border-border-subtle/20 last:border-0" href="/en/contact/">Contact</a><a href="https://github.com/framerslab/agentos" target="_blank" rel="noopener noreferrer" class="px-6 py-4 text-base font-bold text-[var(--color-text-primary)] hover:text-accent-primary hover:bg-accent-primary/5 transition-all duration-300 ease-out flex items-center gap-2 border-b border-border-subtle/20 last:border-0"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-github w-5 h-5" aria-hidden="true"><path d="M15 22v-4a4.8 4.8 0 0 0-1-3.5c3 0 6-2 6-5.5.08-1.25-.27-2.48-1-3.5.28-1.15.28-2.35 0-3.5 0 0-1 0-3 1.5-2.64-.5-5.36-.5-8 0C6 2 5 2 5 2c-.3 1.15-.3 2.35 0 3.5A5.403 5.403 0 0 0 4 9c0 3.5 3 5.5 6 5.5-.39.49-.68 1.05-.85 1.65-.17.6-.22 1.23-.15 1.85v4"></path><path d="M9 18c-4.51 2-5-2-7-2"></path></svg>GitHub</a></nav></div><div class="mx-4 mb-4 mt-2 flex items-center justify-center gap-3 rounded-2xl glass-morphism shadow-modern px-6 py-3 sm:hidden"><button class="flex h-9 w-9 items-center justify-center rounded-lg border border-slate-300 bg-white/80 text-slate-700 shadow-sm transition-theme dark:border-slate-600 dar
71k:bg-slate-800/80 dark:text-slate-200" aria-label="Toggle theme"><div class="h-4 w-4 rounded-full bg-gradient-to-tr from-purple-300 to-pink-300"></div></button><div class="relative"><button class="inline-flex items-center justify-center rounded-lg border border-slate-300 bg-white/80 p-2 text-sm text-slate-700 shadow-sm transition hover:bg-white hover:shadow-md dark:border-slate-600 dark:bg-slate-800/80 dark:text-slate-200 dark:hover:bg-slate-800" aria-label="Select theme" aria-expanded="false"><svg viewBox="0 0 24 24" class="h-5 w-5" aria-hidden="true"><circle cx="12" cy="12" r="4" fill="#06b6d4" opacity="0.9"></circle><circle cx="12" cy="12" r="8" fill="none" stroke="#8b5cf6" stroke-width="1.5" opacity="0.5"></circle><line x1="12" y1="4" x2="12" y2="1" stroke="#06b6d4" stroke-width="1.5" opacity="0.7"></line><line x1="20" y1="12" x2="23" y2="12" stroke="#8b5cf6" stroke-width="1.5" opacity="0.7"></line><line x1="12" y1="20" x2="12" y2="23" stroke="#06b6d4" stroke-width="1.5" opacity="0.7"></line><line x1="4" y1="12" x2="1" y2="12" stroke="#8b5cf6" stroke-width="1.5" opacity="0.7"></line></svg></button></div><div class="relative"><button type="button" class="inline-flex items-center gap-2 px-3 py-2 rounded-xl glass-morphism hover:bg-accent-primary/10 transition-all min-h-[44px] min-w-[44px]" aria-label="Select language" aria-expanded="false" aria-haspopup="true"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-globe w-4 h-4"><circle cx="12" cy="12" r="10"></circle><path d="M12 2a14.5 14.5 0 0 0 0 20 14.5 14.5 0 0 0 0-20"></path><path d="M2 12h20"></path></svg><span class="hidden sm:inline text-sm font-medium">English</span></button></div></div></div></header><!--/$-->
71<script>
72              try {
73                if (window.location.hash) {
74                  var id = window.location.hash.slice(1);
75                  var attempts = 0;
76                  var poll = setInterval(function() {
77                    var el = document.getElementById(id);
78                    if (el) { el.scrollIntoView({ behavior: 'smooth', block: 'start' }); clearInterval(poll); }
79                    if (++attempts >= 20) clearInterval(poll);
80                  }, 200);
81                }
82              } catch(e) {}
83            </script>
83<main id="main-content" tabindex="-1" class="focus:outline-none scroll-mt-24"><!--$--><main id="main-content" class="relative overflow-x-clip bg-[var(--color-background-primary)] text-[var(--color-text-primary)]">
83<script type="application/ld+json">{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://agentos.sh"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://agentos.sh/en/blog/"}]}</script>
83<div class="absolute inset-0 pointer-events-none"><div class="absolute inset-0 bg-gradient-to-br from-purple-500/10 via-transparent to-blue-500/10"></div></div><div class="relative z-10 max-w-6xl mx-auto px-4 sm:px-6 lg:px-8 py-24"><header class="mb-16 text-center"><h1 class="text-4xl sm:text-5xl font-bold mb-4 text-[var(--color-text-primary)]">Blog</h1><p class="text-lg text-[var(--color-text-secondary)] max-w-2xl mx-auto">Technical articles, framework comparisons, and tutorials for building production AI agents with TypeScript.</p></header><section aria-labelledby="featured-heading" class="mb-16"><div class="mb-6 flex items-center gap-3"><span aria-hidden="true" class="block h-[2px] w-10 rounded-full" style="background:linear-gradient(90deg, hsl(180, 95%, 60%), hsl(270, 85%, 65%))"></span><h2 id="featured-heading" class="text-[11px] font-bold uppercase tracking-[0.2em] text-[var(--color-text-muted)]">Featured</h2></div><div class="grid gap-6 grid-cols-1 md:grid-cols-2"><article class="group"><a class="relative flex h-full flex-col overflow-hidden rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-xl" href="/en/blog/paracosm-launch/"><div class="relative aspect-[16/9] w-full overflow-hidden bg-[var(--color-background-tertiary)]"><img src="/img/blog/paracosm/sim-poster.jpg" alt="Paracosm: A structured world-model engine for LLM agent swarms" class="h-full w-full object-cover transition-transform duration-500 group-hover:scale-[1.02]" aria-hidden="true"/><div aria-hidden="true" class="pointer-events-none absolute inset-0" style="background:linear-gradient(180deg, transparent 50%, hsla(220, 30%, 8%, 0.45))"></div></div><div class="flex flex-1 flex-col p-6 sm:p-8"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dark:text-blue-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Engineering</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>May 5, 2026</span></div><h3 class="mb-3 text-2xl font-bold leading-tight text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-3xl">Paracosm: A structured world-model engine for LLM agent swarms</h3><p class="mb-5 line-clamp-4 text-[15px] leading-relaxed text-[var(--color-text-secondary)]">Paracosm is an open-source TypeScript engine that compiles a natural-language scenario into a typed deterministic world and runs HEXACO-typed AI agents through it. Same world, same kernel, same seed, divergent events and futures. This is what it does, how it works, and how I was inspired by interactive fiction, immersive sims, and game design mechanics to build it.</p><span class="mt-auto inline-flex items-center gap-1.5 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-3">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></div></a></article><article class="group"><a class="relative flex h-full flex-col overflow-hidden rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-xl" href="/en/blog/agentos-memory-sota-longmemeval/"><div class="relative aspect-[16/9] w-full overflow-hidden bg-[var(--color-background-tertiary)]"><img src="/img/blog/og/agentos-memory-sota-longmemeval.png" alt="AgentOS beats Mastra on LongMemEval-S at gpt-4o: 85.6% vs 84.23%" class="h-full w-full object-cover transition-transform duration-500 group-hover:scale-[1.02]" aria-hidden="true"/><div aria-hidden="true" class="pointer-events-none absolute inset-0" style="background:linear-gradient(180deg, transparent 50%, hsla(220, 30%, 8%, 0.45))"></div></div><div class="flex flex-1 flex-col p-6 sm:p-8"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dar
83k:text-blue-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Engineering</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 29, 2026</span></div><h3 class="mb-3 text-2xl font-bold leading-tight text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-3xl">AgentOS beats Mastra on LongMemEval-S at gpt-4o: 85.6% vs 84.23%</h3><p class="mb-5 line-clamp-4 text-[15px] leading-relaxed text-[var(--color-text-secondary)]">Plus 70.2% on the 1.5M-token M variant — the only open-source memory lib above 65% on M with reproducible methodology.</p><span class="mt-auto inline-flex items-center gap-1.5 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-3">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></div></a></article></div></section><section aria-labelledby="all-posts-heading"><div class="mb-6 flex items-center gap-3"><span aria-hidden="true" class="block h-[2px] w-10 rounded-full bg-[var(--color-border-primary)]"></span><h2 id="all-posts-heading" class="text-[11px] font-bold uppercase tracking-[0.2em] text-[var(--color-text-muted)]">All posts</h2></div><div id="blog-list-top" class="space-y-6 max-w-4xl mx-auto scroll-mt-24"><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/memory-benchmark-transparency-audit/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dark:text-blue-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Engineering</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 24, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">Memory Benchmark Transparency: Why LongMemEval and LOCOMO Numbers Don&#x27;t Compare</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">An audit of how Mem0, Mastra, Supermemory, Zep, EmergenceMem, and Letta publish their LongMemEval and LOCOMO numbers. LOCOMO has a 6.4% wrong answer key and a judge that accepts 63% of wrong answers. Dhravya posts 99% as a gaming demonstration. Mastra claims 95% at a different (cheaper) answer LLM than the paper specifies. The transparency stack behind every honest comparison.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/inside-mars-genesis-ai-colony-simulation/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dar
83k:text-blue-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Engineering</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 17, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">Inside Mars Genesis: A HEXACO AI Agent Case Study in Emergent Simulation</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">Mars Genesis is a six-turn, thirty-colonist AI civilization simulation where two commanders with opposing HEXACO personality profiles run the same seed and produce measurably different outcomes. This is the end-to-end case study: how the forge-and-reuse economy, personality-weighted memory, and cognitive substrate combine so the Visionary&#x27;s toolbox diverges from the Engineer&#x27;s while both share identical starting conditions.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/ai-companion-case-study-wilds/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-gray-500/10 text-gray-600 dark:text-gray-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Case Study</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 14, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">The AI That Remembers Your Name, Sends You Memes, and Knows When You&#x27;re Lying</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">A user asks their AI companion to describe a GIF it just sent. The companion calls a vision tool, looks at the image, and responds with what it sees. None of this is scripted. Here&#x27;s how AgentOS makes it possible.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/building-ai-companions-agentic-tools/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-orange-500/10 text-orange-600 dar
83k:text-orange-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Tutorial</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 14, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">Building AI Companions with Agentic Tools, Memory, and Personality in TypeScript</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">How to build an AI companion that remembers conversations, sends GIFs, analyzes images, and develops a personality over time. Complete TypeScript implementation with AgentOS.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/adaptive-vs-emergent/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dark:text-blue-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Engineering</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 12, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">Adaptive vs. Emergent Intelligence: Two Types of AI Agent Behavior</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">Adaptive intelligence modifies behavior from feedback. Emergent intelligence produces complex behavior from multi-agent interaction. AgentOS implements both with distinct runtime primitives and neuroscience-grounded memory.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/agentos-vs-langgraph-vs-crewai/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-purple-500/10 text-purple-600 dar
83k:text-purple-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Comparison</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 12, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">AgentOS vs LangGraph vs CrewAI vs Mastra vs VoltAgent: AI Agent Frameworks Compared (2026)</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">A side-by-side comparison of the six leading AI agent frameworks in 2026: features, architecture, code examples, and when to use each. Updated with VoltAgent, OpenAI Agents SDK, and Google ADK.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/announcing-agentos/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-green-500/10 text-green-600 dark:text-green-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Announcements</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 12, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">Announcing AgentOS: Open-Source TypeScript AI Agent Runtime</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">AgentOS is an open-source TypeScript runtime for building AI agents with cognitive memory, HEXACO personality, multi-agent orchestration, runtime tool forging, 37 channel adapters, and 11 LLM providers. Apache 2.0 licensed.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article><article class="group"><a class="block rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] p-6 transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-lg sm:p-8" href="/en/blog/build-typescript-ai-agent-5-minutes/"><div class="mb-3 flex items-center gap-3 text-sm"><span class="inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-orange-500/10 text-orange-600 dar
83k:text-orange-400"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-tag h-3 w-3" aria-hidden="true"><path d="M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z"></path><circle cx="7.5" cy="7.5" r=".5" fill="currentColor"></circle></svg>Tutorial</span><span class="flex items-center gap-1 text-[var(--color-text-muted)]"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-calendar h-3.5 w-3.5" aria-hidden="true"><path d="M8 2v4"></path><path d="M16 2v4"></path><rect width="18" height="18" x="3" y="4" rx="2"></rect><path d="M3 10h18"></path></svg>April 12, 2026</span></div><h2 class="mb-2 text-xl font-bold text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-2xl">Build a TypeScript AI Agent Runtime in 5 Minutes (AgentOS Quickstart)</h2><p class="mb-4 leading-relaxed text-[var(--color-text-secondary)]">From npm install to a working agent with cognitive memory, HEXACO personality, tools, and guardrails. Five steps, under 50 lines of TypeScript. Runtime-anchored quickstart with inline citations.</p><span class="inline-flex items-center gap-1 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-2">Read article<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-right h-4 w-4" aria-hidden="true"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></article></div></section><section class="mx-auto w-full max-w-6xl px-4 sm:px-6 lg:px-8 py-12 mt-16" aria-label="Join the Wilds AI Discord — official community for AgentOS and Paracosm"><a target="_blank" rel="noopener noreferrer" class="group flex items-center gap-4 roun
83ded-xl border border-[var(--color-border-primary)] bg-[var(--color-background-secondary)]/60 p-6 sm:p-8 backdrop-blur-sm transition-all hover:border-[var(--color-accent-primary)] hover:bg-[var(--color-accent-primary)]/5" href="https://wilds.ai/discord"><img src="/logos/wilds-ai-icon.png" alt="Wilds AI" width="64" height="64" loading="lazy" decoding="async" class="flex-shrink-0 rounded-lg h-16 w-16"/><div class="flex-1 min-w-0"><div class="font-semibold text-[var(--color-text-primary)] text-lg sm:text-xl">Join the Wilds AI Discord</div><div class="text-[var(--color-text-secondary)] text-sm sm:text-base mt-1">Official community for AgentOS and Paracosm support, developer onboarding, and questions.</div></div><span class="flex-shrink-0 inline-flex items-center gap-1.5 font-semibold text-[var(--color-accent-primary)] transition-transform group-hover:translate-x-0.5 text-base" aria-hidden="true">Join<svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2.5" stroke-linecap="round" stroke-linejoin="round" class="h-4 w-4"><path d="M5 12h14"></path><path d="m12 5 7 7-7 7"></path></svg></span></a></section></div></main><!--/$--></main><button type="button" aria-label="Scroll back to top" class="fixed bottom-6 right-6 z-[110] inline-flex h-12 w-12 items-center justify-center rounded-full border border-[var(--color-border-subtle)] bg-[var(--color-background-glass)] backdrop-blur-xl text-[var(--color-text-primary)] shadow-lg shadow-[var(--color-accent-primary)]/10 transition-all duration-[var(--duration-smooth)] hover:-translate-y-1 hover:shadow-xl hover:border-[var(--color-border-interactive)] hover:bg-[var(--color-accent-primary)]/10 focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-[var(--color-accent-primary)] opacity-0 pointer-events-none"><svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-arrow-up h-5 w-5"><path d="m5 12 7-7 7 7"></path><path d="M12 19V5"></path></svg></button><footer class="border-t border-purple-200/30 dark:border-purple-500/20 bg-gradient-to-br from-white/90 via-purple-50/30 to-pink-50/30 dark:from-black/80 dark:via-purple-950/40 dark:to-pink-950/40 backdrop-blur-lg py-12"><div class="mx-auto w-full max-w-6xl px-6"><div class="grid md:grid-cols-4 gap-8 mb-8"><div><h3 class="font-bold text-text-primary mb-3">AgentOS</h3><p class="text-sm text-text-secondary mb-3">Open-source TypeScript AI agent runtime by Frame.dev</p><div class="flex gap-3"><a href="https://github.com/framerslab" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors" aria-label="FramersLab on GitHub"><svg class="w-5 h-5" fill="currentColor" viewBox="0 0 24 24"><path d="M12 0c-6.626 0-12 5.373-12 12 0 5.302 3.438 9.8 8.207 11.387.599.111.793-.261.793-.577v-2.234c-3.338.726-4.033-1.416-4.033-1.416-.546-1.387-1.333-1.756-1.333-1.756-1.089-.745.083-.729.083-.729 1.205.084 1.839 1.237 1.839 1.237 1.07 1.834 2.807 1.304 3.492.997.107-.775.418-1.305.762-1.604-2.665-.305-5.467-1.334-5.467-5.931 0-1.311.469-2.381 1.236-3.221-.124-.303-.535-1.524.117-3.176 0 0 1.008-.322 3.301 1.23.957-.266 1.983-.399 3.003-.404 1.02.005 2.047.138 3.006.404 2.291-1.552 3.297-1.23 3.297-1.23.653 1.653.242 2.874.118 3.176.77.84 1.235 1.911 1.235 3.221 0 4.609-2.807 5.624-5.479 5.921.43.372.823 1.102.823 2.222v3.293c0 .319.192.694.801.576 4.765-1.589 8.199-6.086 8.199-11.386 0-6.627-5.373-12-12-12z"></path></svg><span class="sr-only">GitHub</span></a><a href="https://www.linkedin.com/company/framerslab" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors" aria-label="Frame on LinkedIn"><svg class="w-5 h-5" fill="currentColor" viewBox="0 0 24 24"><path d="M20.447 20.452h-3.554v-5.569c0-1.328-.027-3.037-1.852-3.037-1.853 0-2.136 1.445-2.136 2.939v5.667H9.351V9h3.414v1.561h.046c.477-.9 1.637-1.85 3.37-1.85 3.601 0 4.267 2.37 4.267 5.455v6.286zM5.337 7.433c-1.144 0-2.063-.926-2.063-2.065 0-1.138.92-2.063 2.063-2.063 1.14 0 2.064.925 2.064 2.063 0 1.139-.925 2.065-2.064 2.065zm1.782 13.019H3.555V9h3.564v11.452zM22.225 0H1.771C.792 0 0 .774 0 1.729v20.542C0 23.227.792 24 1.771 24h20.451C23.2 24 24 23.227 24 22.271V1.729C24 .774 23.2 0 22.222 0h.003z"></path></svg><span class="sr-only">LinkedIn</span></a><a href="https://twitter.com/framerslab" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors" aria-label="FramersLab on Twitter"><svg class="w-5 h-5" fill="currentColor" viewBox="0 0 24 24">
83<path d="M23.953 4.57a10 10 0 01-2.825.775 4.958 4.958 0 002.163-2.723c-.951.555-2.005.959-3.127 1.184a4.92 4.92 0 00-8.384 4.482C7.69 8.095 4.067 6.13 1.64 3.162a4.822 4.822 0 00-.666 2.475c0 1.71.87 3.213 2.188 4.096a4.904 4.904 0 01-2.228-.616v.06a4.923 4.923 0 003.946 4.827 4.996 4.996 0 01-2.212.085 4.936 4.936 0 004.604 3.417 9.867 9.867 0 01-6.102 2.105c-.39 0-.779-.023-1.17-.067a13.995 13.995 0 007.557 2.209c9.053 0 13.998-7.496 13.998-13.985 0-.21 0-.42-.015-.63A9.935 9.935 0 0024 4.59z"></path></svg><span class="sr-only">Twitter</span></a></div></div><div><h3 class="font-bold text-text-primary mb-3">Resources</h3><ul class="space-y-2 text-sm"><li><a href="https://docs.agentos.sh" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Docs (Guides)</a></li><li><a href="https://docs.agentos.sh/api" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">API Reference (TypeDoc/TSDoc)</a></li><li><a href="/en/guides" class="text-text-secondary hover:text-accent-primary transition-colors">Guides</a></li><li><a href="/en/#code" class="text-text-secondary hover:text-accent-primary transition-colors">Examples</a></li><li><a href="https://github.com/framerslab/agentos/releases" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Release Notes</a></li><li><a href="https://github.com/framerslab/agentos" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">GitHub</a></li><li><a href="https://wilds.ai/discord" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Discord</a></li></ul></div><div><h3 class="font-bold text-text-primary mb-3">Company</h3><ul class="space-y-2 text-sm"><li><a href="https://frame.dev" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Frame.dev</a></li><li><a href="https://wilds.ai" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Wilds.ai</a></li><li><a href="https://paracosm.agentos.sh" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Paracosm</a></li><li><a href="https://wunderland.sh" target="_blank" rel="noopener noreferrer" class="text-text-secondary hover:text-accent-primary transition-colors">Wunderland</a></li><li><a href="/en/about" class="text-text-secondary hover:text-accent-primary transition-colors">About</a></li><li><a href="/en/faq" class="text-text-secondary hover:text-accent-primary transition-colors">FAQ</a></li><li><a href="/en/contact" class="text-text-secondary hover:text-accent-primary transition-colors">Contact</a></li></ul></div><div><h3 class="font-bold text-text-primary mb-3">Legal</h3><ul class="space-y-2 text-sm"><li><a href="/en/legal/terms" class="text-text-secondary hover:text-accent-primary transition-colors">Terms</a></li><li><a href="/en/legal/privacy" class="text-text-secondary hover:text-accent-primary transition-colors">Privacy</a></li><li><a href="/en/legal/security" class="text-text-secondary hover:text-accent-primary transition-colors">Security</a></li><li><a href="/en/legal/cookies" class="text-text-secondary hover:text-accent-primary transition-colors">Cookies</a></li><li><span class="text-text-secondary">Apache 2.0 (core) + MIT (agents, extensions, guardrails)</span></li></ul></div></div><div class="pt-6 border-t border-purple-200/30 dark:border-purple-500/20 text-center"><p class="text-text-secondary">© 2026 Frame.dev. All rights reserved.</p><p class="text-text-secondary text-sm mt-2">Frame is a member of the <a href="https://deepgram.com/startups" target="_blank" rel="noopener noreferrer" class="text-accent-primary hover:underline">Deepgram Startup Program</a> and the <a href="https://www.nvidia.com/en-us/startups/" target="_blank" rel="noopener noreferrer" class="text-accent-primary hover:underline">NVIDIA Inception Program</a></p></div></div></footer><!--/$-->
83<script src="/_next/static/chunks/webpack-bf5df1ec8c5aa8d1.js" async=""></script>
83<script>(self.__next_f=self.__next_f||[]).push([0]);self.__next_f.push([2,null])</script>
83<script>self.__next_f.push([1,"1:HL[\"/_next/static/media/e4af272ccee01ff0-s.p.woff2\",\"font\",{\"crossOrigin\":\"\",\"type\":\"font/woff2\"}]\n2:HL[\"/_next/static/css/34251478a1fd3bc2.css\",\"style\"]\n3:HL[\"/_next/static/css/62f1ca5a4f071d84.css\",\"style\"]\n4:HL[\"/_next/static/css/96b04247e785c9c6.css\",\"style\"]\n"])</script>
83<script>self.__next_f.push([1,"5:I[12433,[],\"\"]\n8:I[90176,[],\"\"]\na:I[39414,[],\"\"]\nc:I[86575,[\"2439\",\"static/chunks/app/%5Blocale%5D/loading-547537401f418075.js\"],\"default\"]\nf:I[93683,[\"3185\",\"static/chunks/app/layout-991763ab09c81468.js\"],\"default\"]\n10:I[57690,[\"8555\",\"static/chunks/app/loading-7b32c3a4f4ce7f0e.js\"],\"default\"]\n12:I[78528,[],\"\"]\n9:[\"locale\",\"en\",\"d\"]\nd:T866,"])</script>
83<script>self.__next_f.push([1,"\n              (function() {\n                try {\n                  var THEME_VERSION = '2026-04-29-twilight-default';\n                  var modeKey = 'agentos-mode';\n                  var themeKey = 'agentos-theme';\n                  var versionKey = 'agentos-theme-version';\n\n                  var ver = localStorage.getItem(versionKey);\n                  if (ver !== THEME_VERSION) {\n                    localStorage.setItem(versionKey, THEME_VERSION);\n                    localStorage.removeItem(themeKey);\n                  }\n\n                  var mode = localStorage.getItem(modeKey);\n                  if (!mode || mode === 'system') {\n                    mode = window.matchMedia \u0026\u0026 window.matchMedia('(prefers-color-scheme: dark)').matches ? 'dark' : 'light';\n                  }\n                  if (mode === 'dark') {\n                    document.documentElement.classList.add('dark');\n                    document.documentElement.style.colorScheme = 'dark';\n                  } else {\n                    document.documentElement.classList.remove('dark');\n                    document.documentElement.style.colorScheme = 'light';\n                  }\n\n                  var theme = localStorage.getItem(themeKey) || 'twilight-neo';\n                  document.documentElement.setAttribute('data-theme', theme);\n\n                  // BenchmarkBanner dismissal — sync-check before paint so\n                  // returning visitors don't see the banner flash in then\n                  // out (the React useState/useEffect path renders visible\n                  // → measures localStorage → hides, producing a ~70px CLS\n                  // shift Lighthouse attributed 0.16 to on mobile). Keep\n                  // the storage key in sync with the BenchmarkBanner\n                  // STORAGE_KEY constant.\n                  if (localStorage.getItem('agentos-banner-2026-04-29') === 'dismissed') {\n                    document.documentElement.setAttribute('data-banner-dismissed', 'true');\n                  }\n                } catch (e) { /* swallow: localStorage may be unavailable in some embeds */ }\n              })();\n            "])</script>
83<script>self.__next_f.push([1,"e:T774,\n          * { box-sizing: border-box; }\n          html, body { width: 100%; margin: 0; padding: 0; }\n          html { height: 100%; overflow-x: clip; }\n          body { min-height: 100vh; overflow-x: clip; }\n\n          .skip-to-content { position: absolute; left: -9999px; z-index: 999; padding: 1rem 1.5rem; background: hsl(250, 95%, 55%); color: white; text-decoration: none; border-radius: 0.5rem; font-weight: 600; }\n          .skip-to-content:focus { left: 1rem; top: 1rem; outline: 2px solid hsl(280, 85%, 60%); outline-offset: 2px; }\n          .skeleton { position: relative; overflow: hidden; border-radius: 0.5rem; }\n          .skeleton::after { content: \"\"; position: absolute; inset: 0; background: linear-gradient(90deg, transparent 0%, rgba(200,200,200,0.3) 50%, transparent 100%); animation: skeleton-shimmer 2s ease-in-out infinite; }\n          @keyframes skeleton-shimmer { 0% { transform: translateX(-100%); } 100% { transform: translateX(100%); } }\n\n          .hero-critical { display: flex; flex-direction: column; justify-content: center; position: relative; min-height: 100vh; overflow: hidden; }\n          .hero-critical \u003e * { position: relative; z-index: 1; }\n\n          /* Banner dismissal — paired with the bootstrap script that sets\n             data-banner-dismissed before paint. The banner's own\n             useState(false) default would otherwise render visible then\n             vanish on hydration, shifting the hero down by ~70px. */\n          html[data-banner-dismissed=\"true\"] [aria-label=\"Latest benchmark result\"] { display: none; }\n\n          @media (prefers-reduced-motion: reduce) {\n            *, *::before, *::after {\n              animation-duration: 0.01ms !important;\n              animation-iteration-count: 1 !important;\n              transition-duration: 0.01ms !important;\n              scroll-behavior: auto !important;\n            }\n          }\n        13:[]\n"])</script>
83<script>self.__next_f.push([1,"0:[\"$\",\"$L5\",null,{\"buildId\":\"r8S8rA69qbtyXVFzweNXH\",\"assetPrefix\":\"\",\"urlParts\":[\"\",\"en\",\"blog\",\"\"],\"initialTree\":[\"\",{\"children\":[[\"locale\",\"en\",\"d\"],{\"children\":[\"blog\",{\"children\":[\"__PAGE__?{\\\"locale\\\":\\\"en\\\"}\",{}]}]}]},\"$undefined\",\"$undefined\",true],\"initialSeedData\":[\"\",{\"children\":[[\"locale\",\"en\",\"d\"],{\"children\":[\"blog\",{\"children\":[\"__PAGE__\",{},[[\"$L6\",\"$L7\",null],null],null]},[null,[\"$\",\"$L8\",null,{\"parallelRouterKey\":\"children\",\"segmentPath\":[\"children\",\"$9\",\"children\",\"blog\",\"children\"],\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$La\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"notFoundStyles\":\"$undefined\"}]],null]},[[null,\"$Lb\"],null],[[\"$\",\"$Lc\",null,{}],[],[]]]},[[[[\"$\",\"link\",\"0\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/css/34251478a1fd3bc2.css\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\"}],[\"$\",\"link\",\"1\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/css/62f1ca5a4f071d84.css\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\"}],[\"$\",\"link\",\"2\",{\"rel\":\"stylesheet\",\"href\":\"/_next/static/css/96b04247e785c9c6.css\",\"precedence\":\"next\",\"crossOrigin\":\"$undefined\"}]],[\"$\",\"html\",null,{\"lang\":\"en\",\"suppressHydrationWarning\":true,\"children\":[[\"$\",\"head\",null,{\"children\":[[\"$\",\"link\",null,{\"rel\":\"icon\",\"href\":\"/favicon.ico\"}],[\"$\",\"link\",null,{\"rel\":\"icon\",\"type\":\"image/svg+xml\",\"href\":\"/favicon.svg\"}],[\"$\",\"script\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"$d\"}}],[\"$\",\"style\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"$e\"}}]]}],[\"$\",\"body\",null,{\"suppressHydrationWarning\":true,\"className\":\"__variable_7748ca grainy min-h-screen antialiased transition-theme bg-background-primary text-text-primary\",\"children\":[[\"$\",\"$Lf\",null,{}],[\"$\",\"$L8\",null,{\"parallelRouterKey\":\"children\",\"segmentPath\":[\"children\"],\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$La\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":[\"$\",\"div\",null,{\"style\":{\"minHeight\":\"100vh\",\"display\":\"flex\",\"flexDirection\":\"column\",\"alignItems\":\"center\",\"justifyContent\":\"center\",\"padding\":\"2rem\",\"textAlign\":\"center\",\"background\":\"var(--color-background-primary, hsl(240,20%,3%))\",\"color\":\"var(--color-text-primary, hsl(220,30%,98%))\"},\"children\":[[\"$\",\"h1\",null,{\"style\":{\"fontSize\":\"5rem\",\"fontWeight\":800,\"marginBottom\":\"0.5rem\",\"background\":\"linear-gradient(135deg, hsl(250,100%,70%), hsl(280,100%,65%))\",\"WebkitBackgroundClip\":\"text\",\"WebkitTextFillColor\":\"transparent\",\"backgroundClip\":\"text\"},\"children\":\"404\"}],[\"$\",\"h2\",null,{\"style\":{\"fontSize\":\"1.5rem\",\"fontWeight\":600,\"marginBottom\":\"0.75rem\"},\"children\":\"Page Not Found\"}],[\"$\",\"p\",null,{\"style\":{\"fontSize\":\"1rem\",\"color\":\"var(--color-text-muted, hsl(220,15%,60%))\",\"maxWidth\":480,\"marginBottom\":\"2rem\",\"lineHeight\":1.6},\"children\":\"The page you are looking for does not exist or has been moved.\"}],[\"$\",\"div\",null,{\"style\":{\"display\":\"flex\",\"gap\":\"0.75rem\",\"flexWrap\":\"wrap\",\"justifyContent\":\"center\"},\"children\":[[\"$\",\"a\",null,{\"href\":\"/\",\"style\":{\"padding\":\"0.75rem 1.5rem\",\"borderRadius\":\"0.5rem\",\"background\":\"var(--color-accent-primary, hsl(250,100%,70%))\",\"color\":\"#fff\",\"textDecoration\":\"none\",\"fontWeight\":600},\"children\":\"Go Home\"}],[\"$\",\"a\",null,{\"href\":\"https://docs.agentos.sh\",\"style\":{\"padding\":\"0.75rem 1.5rem\",\"borderRadius\":\"0.5rem\",\"border\":\"1px solid var(--color-border-subtle, hsl(260,20%,18%))\",\"color\":\"var(--color-text-primary, hsl(220,30%,98%))\",\"textDecoration\":\"none\",\"fontWeight\":600},\"children\":\"View Docs\"}],[\"$\",\"a\",null,{\"href\":\"https://github.com/wearetheframers/agentos\",\"style\":{\"padding\":\"0.75rem 1.5rem\",\"borderRadius\":\"0.5rem\",\"border\":\"1px solid var(--color-border-subtle, hsl(2
8360,20%,18%))\",\"color\":\"var(--color-text-primary, hsl(220,30%,98%))\",\"textDecoration\":\"none\",\"fontWeight\":600},\"children\":\"GitHub\"}]]}]]}],\"notFoundStyles\":[]}]]}]]}]],null],[[\"$\",\"$L10\",null,{}],[],[]]],\"couldBeIntercepted\":false,\"initialHead\":[null,\"$L11\"],\"globalErrorComponent\":\"$12\",\"missingSlots\":\"$W13\"}]\n"])</script>
83<script>self.__next_f.push([1,"14:I[69498,[\"326\",\"static/chunks/app/%5Blocale%5D/blog/page-a30ab686764a69d2.js\"],\"\"]\n15:I[21892,[\"326\",\"static/chunks/app/%5Blocale%5D/blog/page-a30ab686764a69d2.js\"],\"PaginatedPosts\"]\n"])</script>
83<script>self.__next_f.push([1,"7:[\"$\",\"main\",null,{\"id\":\"main-content\",\"className\":\"relative overflow-x-clip bg-[var(--color-background-primary)] text-[var(--color-text-primary)]\",\"children\":[[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"BreadcrumbList\\\",\\\"itemListElement\\\":[{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":1,\\\"name\\\":\\\"Home\\\",\\\"item\\\":\\\"https://agentos.sh\\\"},{\\\"@type\\\":\\\"ListItem\\\",\\\"position\\\":2,\\\"name\\\":\\\"Blog\\\",\\\"item\\\":\\\"https://agentos.sh/en/blog/\\\"}]}\"}}],[\"$\",\"div\",null,{\"className\":\"absolute inset-0 pointer-events-none\",\"children\":[\"$\",\"div\",null,{\"className\":\"absolute inset-0 bg-gradient-to-br from-purple-500/10 via-transparent to-blue-500/10\"}]}],[\"$\",\"div\",null,{\"className\":\"relative z-10 max-w-6xl mx-auto px-4 sm:px-6 lg:px-8 py-24\",\"children\":[[\"$\",\"header\",null,{\"className\":\"mb-16 text-center\",\"children\":[[\"$\",\"h1\",null,{\"className\":\"text-4xl sm:text-5xl font-bold mb-4 text-[var(--color-text-primary)]\",\"children\":\"Blog\"}],[\"$\",\"p\",null,{\"className\":\"text-lg text-[var(--color-text-secondary)] max-w-2xl mx-auto\",\"children\":\"Technical articles, framework comparisons, and tutorials for building production AI agents with TypeScript.\"}]]}],[\"$\",\"section\",null,{\"aria-labelledby\":\"featured-heading\",\"className\":\"mb-16\",\"children\":[[\"$\",\"div\",null,{\"className\":\"mb-6 flex items-center gap-3\",\"children\":[[\"$\",\"span\",null,{\"aria-hidden\":true,\"className\":\"block h-[2px] w-10 rounded-full\",\"style\":{\"background\":\"linear-gradient(90deg, hsl(180, 95%, 60%), hsl(270, 85%, 65%))\"}}],[\"$\",\"h2\",null,{\"id\":\"featured-heading\",\"className\":\"text-[11px] font-bold uppercase tracking-[0.2em] text-[var(--color-text-muted)]\",\"children\":\"Featured\"}]]}],[\"$\",\"div\",null,{\"className\":\"grid gap-6 grid-cols-1 md:grid-cols-2\",\"children\":[[\"$\",\"article\",\"paracosm-launch\",{\"className\":\"group\",\"children\":[\"$\",\"$L14\",null,{\"href\":\"/en/blog/paracosm-launch\",\"className\":\"relative flex h-full flex-col overflow-hidden rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-xl\",\"children\":[[\"$\",\"div\",null,{\"className\":\"relative aspect-[16/9] w-full overflow-hidden bg-[var(--color-background-tertiary)]\",\"children\":[[\"$\",\"img\",null,{\"src\":\"/img/blog/paracosm/sim-poster.jpg\",\"alt\":\"Paracosm: A structured world-model engine for LLM agent swarms\",\"className\":\"h-full w-full object-cover transition-transform duration-500 group-hover:scale-[1.02]\",\"aria-hidden\":true}],[\"$\",\"div\",null,{\"aria-hidden\":true,\"className\":\"pointer-events-none absolute inset-0\",\"style\":{\"background\":\"linear-gradient(180deg, transparent 50%, hsla(220, 30%, 8%, 0.45))\"}}]]}],[\"$\",\"div\",null,{\"className\":\"flex flex-1 flex-col p-6 sm:p-8\",\"children\":[[\"$\",\"div\",null,{\"className\":\"mb-3 flex items-center gap-3 text-sm\",\"children\":[[\"$\",\"span\",null,{\"className\":\"inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dark:text-blue-400\",\"children\":[[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":2,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"lucide lucide-tag h-3 w-3\",\"aria-hidden\":\"true\",\"children\":[[\"$\",\"path\",\"vktsd0\",{\"d\":\"M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z\"}],[\"$\",\"circle\",\"kqv944\",{\"cx\":\"7.5\",\"cy\":\"7.5\",\"r\":\".5\",\"fill\":\"currentColor\"}],\"$undefined\"]}],\"Engineering\"]}],[\"$\",\"span\",null,{\"className\":\"flex items-center gap-1 text-[var(--color-text-muted)]\",\"children\":[[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\"
83:24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":2,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"lucide lucide-calendar h-3.5 w-3.5\",\"aria-hidden\":\"true\",\"children\":[[\"$\",\"path\",\"1cmpym\",{\"d\":\"M8 2v4\"}],[\"$\",\"path\",\"4m81vk\",{\"d\":\"M16 2v4\"}],[\"$\",\"rect\",\"1hopcy\",{\"width\":\"18\",\"height\":\"18\",\"x\":\"3\",\"y\":\"4\",\"rx\":\"2\"}],[\"$\",\"path\",\"8toen8\",{\"d\":\"M3 10h18\"}],\"$undefined\"]}],\"May 5, 2026\"]}]]}],[\"$\",\"h3\",null,{\"className\":\"mb-3 text-2xl font-bold leading-tight text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-3xl\",\"children\":\"Paracosm: A structured world-model engine for LLM agent swarms\"}],[\"$\",\"p\",null,{\"className\":\"mb-5 line-clamp-4 text-[15px] leading-relaxed text-[var(--color-text-secondary)]\",\"children\":\"Paracosm is an open-source TypeScript engine that compiles a natural-language scenario into a typed deterministic world and runs HEXACO-typed AI agents through it. Same world, same kernel, same seed, divergent events and futures. This is what it does, how it works, and how I was inspired by interactive fiction, immersive sims, and game design mechanics to build it.\"}],[\"$\",\"span\",null,{\"className\":\"mt-auto inline-flex items-center gap-1.5 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-3\",\"children\":[\"Read article\",[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":2,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"lucide lucide-arrow-right h-4 w-4\",\"aria-hidden\":\"true\",\"children\":[[\"$\",\"path\",\"1ays0h\",{\"d\":\"M5 12h14\"}],[\"$\",\"path\",\"xquz4c\",{\"d\":\"m12 5 7 7-7 7\"}],\"$undefined\"]}]]}]]}]]}]}],[\"$\",\"article\",\"agentos-memory-sota-longmemeval\",{\"className\":\"group\",\"children\":[\"$\",\"$L14\",null,{\"href\":\"/en/blog/agentos-memory-sota-longmemeval\",\"className\":\"relative flex h-full flex-col overflow-hidden rounded-2xl border border-[var(--color-border-subtle)] bg-[var(--color-background-card,var(--color-background-elevated))] transition-all hover:border-[var(--color-text-link)]/40 hover:shadow-xl\",\"children\":[[\"$\",\"div\",null,{\"className\":\"relative aspect-[16/9] w-full overflow-hidden bg-[var(--color-background-tertiary)]\",\"children\":[[\"$\",\"img\",null,{\"src\":\"/img/blog/og/agentos-memory-sota-longmemeval.png\",\"alt\":\"AgentOS beats Mastra on LongMemEval-S at gpt-4o: 85.6% vs 84.23%\",\"className\":\"h-full w-full object-cover transition-transform duration-500 group-hover:scale-[1.02]\",\"aria-hidden\":true}],[\"$\",\"div\",null,{\"aria-hidden\":true,\"className\":\"pointer-events-none absolute inset-0\",\"style\":{\"background\":\"linear-gradient(180deg, transparent 50%, hsla(220, 30%, 8%, 0.45))\"}}]]}],[\"$\",\"div\",null,{\"className\":\"flex flex-1 flex-col p-6 sm:p-8\",\"children\":[[\"$\",\"div\",null,{\"className\":\"mb-3 flex items-center gap-3 text-sm\",\"children\":[[\"$\",\"span\",null,{\"className\":\"inline-flex items-center gap-1 rounded-full px-2.5 py-0.5 text-xs font-semibold bg-blue-500/10 text-blue-600 dark:text-blue-400\",\"children\":[[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":2,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"lucide lucide-tag h-3 w-3\",\"aria-hidden\":\"true\",\"children\":[[\"$\",\"path\",\"vktsd0\",{\"d\":\"M12.586 2.586A2 2 0 0 0 11.172 2H4a2 2 0 0 0-2 2v7.172a2 2 0 0 0 .586 1.414l8.704 8.704a2.426 2.426 0 0 0 3.42 0l6.58-6.58a2.426 2.426 0 0 0 0-3.42z\"}],[\"$\",\"circle\",\"kqv944\",{\"cx\":\"7.5\",\"cy\":\"7.5\",\"r\":\".5\",\"fill\":\"currentColor\"}],\"$undefined\"]}],\"Engineering\"]}],[\"$\",\"span\",null,{\"className\":\"flex items-center gap-1 text-[var(--color-text-muted)]\",\"children\":[[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\"
83:24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":2,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"lucide lucide-calendar h-3.5 w-3.5\",\"aria-hidden\":\"true\",\"children\":[[\"$\",\"path\",\"1cmpym\",{\"d\":\"M8 2v4\"}],[\"$\",\"path\",\"4m81vk\",{\"d\":\"M16 2v4\"}],[\"$\",\"rect\",\"1hopcy\",{\"width\":\"18\",\"height\":\"18\",\"x\":\"3\",\"y\":\"4\",\"rx\":\"2\"}],[\"$\",\"path\",\"8toen8\",{\"d\":\"M3 10h18\"}],\"$undefined\"]}],\"April 29, 2026\"]}]]}],[\"$\",\"h3\",null,{\"className\":\"mb-3 text-2xl font-bold leading-tight text-[var(--color-text-primary)] transition-colors group-hover:text-[var(--color-text-link)] sm:text-3xl\",\"children\":\"AgentOS beats Mastra on LongMemEval-S at gpt-4o: 85.6% vs 84.23%\"}],[\"$\",\"p\",null,{\"className\":\"mb-5 line-clamp-4 text-[15px] leading-relaxed text-[var(--color-text-secondary)]\",\"children\":\"Plus 70.2% on the 1.5M-token M variant — the only open-source memory lib above 65% on M with reproducible methodology.\"}],[\"$\",\"span\",null,{\"className\":\"mt-auto inline-flex items-center gap-1.5 text-sm font-semibold text-[var(--color-text-link)] transition-all group-hover:gap-3\",\"children\":[\"Read article\",[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":24,\"height\":24,\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":2,\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"lucide lucide-arrow-right h-4 w-4\",\"aria-hidden\":\"true\",\"children\":[[\"$\",\"path\",\"1ays0h\",{\"d\":\"M5 12h14\"}],[\"$\",\"path\",\"xquz4c\",{\"d\":\"m12 5 7 7-7 7\"}],\"$undefined\"]}]]}]]}]]}]}]]}]]}],[\"$\",\"section\",null,{\"aria-labelledby\":\"all-posts-heading\",\"children\":[[\"$\",\"div\",null,{\"className\":\"mb-6 flex items-center gap-3\",\"children\":[[\"$\",\"span\",null,{\"aria-hidden\":true,\"className\":\"block h-[2px] w-10 rounded-full bg-[var(--color-border-primary)]\"}],[\"$\",\"h2\",null,{\"id\":\"all-posts-heading\",\"className\":\"text-[11px] font-bold uppercase tracking-[0.2em] text-[var(--color-text-muted)]\",\"children\":\"All posts\"}]]}],[\"$\",\"$L15\",null,{\"posts\":[{\"slug\":\"memory-benchmark-transparency-audit\",\"title\":\"Memory Benchmark Transparency: Why LongMemEval and LOCOMO Numbers Don't Compare\",\"excerpt\":\"An audit of how Mem0, Mastra, Supermemory, Zep, EmergenceMem, and Letta publish their LongMemEval and LOCOMO numbers. LOCOMO has a 6.4% wrong answer key and a judge that accepts 63% of wrong answers. Dhravya posts 99% as a gaming demonstration. Mastra claims 95% at a different (cheaper) answer LLM than the paper specifies. The transparency stack behind every honest comparison.\",\"date\":\"2026-04-24\",\"category\":\"Engineering\"},{\"slug\":\"inside-mars-genesis-ai-colony-simulation\",\"title\":\"Inside Mars Genesis: A HEXACO AI Agent Case Study in Emergent Simulation\",\"excerpt\":\"Mars Genesis is a six-turn, thirty-colonist AI civilization simulation where two commanders with opposing HEXACO personality profiles run the same seed and produce measurably different outcomes. This is the end-to-end case study: how the forge-and-reuse economy, personality-weighted memory, and cognitive substrate combine so the Visionary's toolbox diverges from the Engineer's while both share identical starting conditions.\",\"date\":\"2026-04-17\",\"category\":\"Engineering\"},{\"slug\":\"ai-companion-case-study-wilds\",\"title\":\"The AI That Remembers Your Name, Sends You Memes, and Knows When You're Lying\",\"excerpt\":\"A user asks their AI companion to describe a GIF it just sent. The companion calls a vision tool, looks at the image, and responds with what it sees. None of this is scripted. Here's how AgentOS makes it possible.\",\"date\":\"2026-04-14\",\"category\":\"Case Study\"},{\"slug\":\"building-ai-companions-agentic-tools\",\"title\":\"Building AI Companions with Agentic Tools, Memory, and Personality in TypeScript\",\"excerpt\":\"How to build an AI companion that remembers conversations, sends GIFs, analyzes images, and develops a personality over time. Complete TypeScript implementation with AgentOS.\",\"date\":\"2026-04-14\",\"category\":\"Tutorial\"},{\"slug\":\"adaptive-vs-emergent\",\"title\":\"Adaptive vs. Emergent Intelligence: Two Types of AI Agent Behavior\",\"excerpt\":\"Adaptive intelligence modifies behavior from feedback. Emergent intelligence produces complex behavior from multi-agent interaction. AgentOS implements both with distinct runtime primitives and neuroscience-grounded memory.\",\"date\":\"2026-04-12\",\"category\":\"Engineering\"},{\"slug\":\"agentos-vs-langgraph-vs-crewai\",\"title\":\"AgentOS vs LangGraph vs CrewAI vs Mastra vs VoltAgent: AI Agent Frameworks Compared (2026)\",\"excerpt\":\"A side-by-side comparison of the six leading AI agent frameworks in 2026: features, architecture, code examples, and when to use each. Updated with VoltAgent, OpenAI Agents SDK, and Google ADK.\",\"date\":\"2026-04-12\",\"category\":\"Comparison\"}
83,{\"slug\":\"announcing-agentos\",\"title\":\"Announcing AgentOS: Open-Source TypeScript AI Agent Runtime\",\"excerpt\":\"AgentOS is an open-source TypeScript runtime for building AI agents with cognitive memory, HEXACO personality, multi-agent orchestration, runtime tool forging, 37 channel adapters, and 11 LLM providers. Apache 2.0 licensed.\",\"date\":\"2026-04-12\",\"category\":\"Announcements\"},{\"slug\":\"build-typescript-ai-agent-5-minutes\",\"title\":\"Build a TypeScript AI Agent Runtime in 5 Minutes (AgentOS Quickstart)\",\"excerpt\":\"From npm install to a working agent with cognitive memory, HEXACO personality, tools, and guardrails. Five steps, under 50 lines of TypeScript. Runtime-anchored quickstart with inline citations.\",\"date\":\"2026-04-12\",\"category\":\"Tutorial\"}],\"locale\":\"en\",\"pageSize\":10}]]}],false,[\"$\",\"section\",null,{\"className\":\"mx-auto w-full max-w-6xl px-4 sm:px-6 lg:px-8 py-12 mt-16\",\"aria-label\":\"Join the Wilds AI Discord — official community for AgentOS and Paracosm\",\"children\":[\"$\",\"$L14\",null,{\"href\":\"https://wilds.ai/discord\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"group flex items-center gap-4 roun
83ded-xl border border-[var(--color-border-primary)] bg-[var(--color-background-secondary)]/60 p-6 sm:p-8 backdrop-blur-sm transition-all hover:border-[var(--color-accent-primary)] hover:bg-[var(--color-accent-primary)]/5\",\"children\":[[\"$\",\"img\",null,{\"src\":\"/logos/wilds-ai-icon.png\",\"alt\":\"Wilds AI\",\"width\":64,\"height\":64,\"loading\":\"lazy\",\"decoding\":\"async\",\"className\":\"flex-shrink-0 rounded-lg h-16 w-16\"}],[\"$\",\"div\",null,{\"className\":\"flex-1 min-w-0\",\"children\":[[\"$\",\"div\",null,{\"className\":\"font-semibold text-[var(--color-text-primary)] text-lg sm:text-xl\",\"children\":\"Join the Wilds AI Discord\"}],[\"$\",\"div\",null,{\"className\":\"text-[var(--color-text-secondary)] text-sm sm:text-base mt-1\",\"children\":\"Official community for AgentOS and Paracosm support, developer onboarding, and questions.\"}]]}],[\"$\",\"span\",null,{\"className\":\"flex-shrink-0 inline-flex items-center gap-1.5 font-semibold text-[var(--color-accent-primary)] transition-transform group-hover:translate-x-0.5 text-base\",\"aria-hidden\":\"true\",\"children\":[\"Join\",[\"$\",\"svg\",null,{\"xmlns\":\"http://www.w3.org/2000/svg\",\"width\":\"16\",\"height\":\"16\",\"viewBox\":\"0 0 24 24\",\"fill\":\"none\",\"stroke\":\"currentColor\",\"strokeWidth\":\"2.5\",\"strokeLinecap\":\"round\",\"strokeLinejoin\":\"round\",\"className\":\"h-4 w-4\",\"children\":[[\"$\",\"path\",null,{\"d\":\"M5 12h14\"}],[\"$\",\"path\",null,{\"d\":\"m12 5 7 7-7 7\"}]]}]]}]]}]}]]}]]}]\n"])</script>
83<script>self.__next_f.push([1,"11:[[\"$\",\"meta\",\"0\",{\"name\":\"viewport\",\"content\":\"width=device-width, initial-scale=1, maximum-scale=5\"}],[\"$\",\"meta\",\"1\",{\"charSet\":\"utf-8\"}],[\"$\",\"title\",\"2\",{\"children\":\"AgentOS Blog | TypeScript AI Agent Framework Tutorials \u0026 Benchmarks\"}],[\"$\",\"meta\",\"3\",{\"name\":\"description\",\"content\":\"Technical articles, framework comparisons, tutorials, and engineering insights from the AgentOS team. Learn how to build production AI agents with TypeScript.\"}],[\"$\",\"link\",\"4\",{\"rel\":\"author\",\"href\":\"https://frame.dev\"}],[\"$\",\"meta\",\"5\",{\"name\":\"author\",\"content\":\"Frame.dev\"}],[\"$\",\"meta\",\"6\",{\"name\":\"keywords\",\"content\":\"AgentOS,agentos,AI agent framework,TypeScript AI agent framework,AI agent SDK TypeScript,open source AI agent framework,build AI agents,build AI agents TypeScript,AI agent engineering platform,autonomous AI agents,production-ready AI agents,multi-agent orchestration,multi-agent system,agentic AI framework,agentic workflows,agent simulation framework,AI agent simulation,cognitive AI agent,AI agent memory system,adaptive intelligence,emergent behaviors,AI guardrails,prompt injection defense,cognitive memory,multimodal RAG,RAG pipeline TypeScript,voice AI agents,AI agent orchestration patterns,graph orchestration,AI workflow automation,self-hosted AI agents,agent runtime,runtime tool forging,self-improving agents,type-safe AI agents,LangGraph alternative,CrewAI alternative,AutoGen alternative,Mastra alternative,VoltAgent alternative,Vercel AI SDK alternative,OpenAI Agents SDK alternative\"}],[\"$\",\"meta\",\"7\",{\"name\":\"creator\",\"content\":\"Frame.dev\"}],[\"$\",\"meta\",\"8\",{\"name\":\"publisher\",\"content\":\"Frame.dev\"}],[\"$\",\"meta\",\"9\",{\"name\":\"robots\",\"content\":\"index, follow\"}],[\"$\",\"meta\",\"10\",{\"name\":\"googlebot\",\"content\":\"index, follow, max-video-preview:-1, max-image-preview:large, max-snippet:-1\"}],[\"$\",\"link\",\"11\",{\"rel\":\"canonical\",\"href\":\"https://agentos.sh/blog/\"}],[\"$\",\"link\",\"12\",{\"rel\":\"alternate\",\"hrefLang\":\"en\",\"href\":\"https://agentos.sh/blog/\"}],[\"$\",\"link\",\"13\",{\"rel\":\"alternate\",\"hrefLang\":\"es\",\"href\":\"https://agentos.sh/es/blog/\"}],[\"$\",\"link\",\"14\",{\"rel\":\"alternate\",\"hrefLang\":\"fr\",\"href\":\"https://agentos.sh/fr/blog/\"}],[\"$\",\"link\",\"15\",{\"rel\":\"alternate\",\"hrefLang\":\"de\",\"href\":\"https://agentos.sh/de/blog/\"}],[\"$\",\"link\",\"16\",{\"rel\":\"alternate\",\"hrefLang\":\"pt\",\"href\":\"https://agentos.sh/pt/blog/\"}],[\"$\",\"link\",\"17\",{\"rel\":\"alternate\",\"hrefLang\":\"ja\",\"href\":\"https://agentos.sh/ja/blog/\"}],[\"$\",\"link\",\"18\",{\"rel\":\"alternate\",\"hrefLang\":\"ko\",\"href\":\"https://agentos.sh/ko/blog/\"}],[\"$\",\"link\",\"19\",{\"rel\":\"alternate\",\"hrefLang\":\"zh\",\"href\":\"https://agentos.sh/zh/blog/\"}],[\"$\",\"link\",\"20\",{\"rel\":\"alternate\",\"hrefLang\":\"x-default\",\"href\":\"https://agentos.sh/blog/\"}],[\"$\",\"meta\",\"21\",{\"property\":\"og:title\",\"content\":\"AgentOS Blog | TypeScript AI Agent Framework Tutorials \u0026 Benchmarks\"}],[\"$\",\"meta\",\"22\",{\"property\":\"og:description\",\"content\":\"Technical articles, framework comparisons, tutorials, and engineering insights from the AgentOS team.\"}],[\"$\",\"meta\",\"23\",{\"property\":\"og:url\",\"content\":\"https://agentos.sh/blog/\"}],[\"$\",\"meta\",\"24\",{\"property\":\"og:site_name\",\"content\":\"AgentOS\"}],[\"$\",\"meta\",\"25\",{\"property\":\"og:image\",\"content\":\"https://agentos.sh/og-image-v2.png\"}],[\"$\",\"meta\",\"26\",{\"property\":\"og:type\",\"content\":\"website\"}],[\"$\",\"meta\",\"27\",{\"name\":\"twitter:card\",\"content\":\"summary_large_image\"}],[\"$\",\"meta\",\"28\",{\"name\":\"twitter:title\",\"content\":\"AgentOS Blog | TypeScript AI Agent Framework Tutorials \u0026 Benchmarks\"}],[\"$\",\"meta\",\"29\",{\"name\":\"twitter:description\",\"content\":\"Technical articles, framework comparisons, and tutorials for building AI agents with TypeScript.\"}],[\"$\",\"meta\",\"30\",{\"name\":\"twitter:image\",\"content\":\"https://agentos.sh/og-image-v2.png\"}],[\"$\",\"link\",\"31\",{\"rel\":\"shortcut icon\",\"href\":\"/favicon.svg\"}],[\"$\",\"link\",\"32\",{\"rel\":\"icon\",\"href\":\"/favicon.svg\",\"type\":\"image/svg+xml\"}],[\"$\",\"link\",\"33\",{\"rel\":\"icon\",\"href\":\"/icon-128.png\",\"sizes\":\"128x128\",\"type\":\"image/png\"}],[\"$\",\"link\",\"34\",{\"rel\":\"icon\",\"href\":\"/icon-256.png\",\"sizes\":\"256x256\",\"type\":\"image/png\"}],[\"$\",\"link\",\"35\",{\"rel\":\"icon\",\"href\":\"/icon-512.png\",\"sizes\":\"512x512\",\"type\":\"image/png\"}],[\"$\",\"link\",\"36\",{\"rel\":\"apple-touch-icon\",\"href\":\"/icon-256.png\",\"sizes\":\"256x256\",\"type\":\"image/png\"}],[\"$\",\"meta\",\"37\",{\"name\":\"next-size-adjust\"}]]\n"])</script>
83<script>self.__next_f.push([1,"6:null\n"])</script>
83<script>self.__next_f.push([1,"16:I[49577,[\"7586\",\"static/chunks/7586-203f1916d23f2407.js\",\"9079\",\"static/chunks/9079-00ef10f8b11f4d94.js\",\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"4400\",\"static/chunks/4400-4ff185dce52c1c17.js\",\"7531\",\"static/chunks/7531-64596f3089411931.js\",\"5061\",\"static/chunks/app/%5Blocale%5D/page-45ab204874fc5d47.js\"],\"default\"]\n19:I[28607,[\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"1203\",\"static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js\"],\"ThemeProvider\"]\n1b:\"$Sreact.suspense\"\n1c:I[44377,[\"7586\",\"static/chunks/7586-203f1916d23f2407.js\",\"9079\",\"static/chunks/9079-00ef10f8b11f4d94.js\",\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"4400\",\"static/chunks/4400-4ff185dce52c1c17.js\",\"7531\",\"static/chunks/7531-64596f3089411931.js\",\"5061\",\"static/chunks/app/%5Blocale%5D/page-45ab204874fc5d47.js\"],\"BailoutToCSR\"]\n1d:I[70656,[\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"1203\",\"static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js\"],\"GoogleAnalytics\",1]\n1e:I[95666,[\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"1203\",\"static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js\"],\"MicrosoftClarity\",1]\n1f:I[92082,[\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"1203\",\"static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js\"],\"CookieConsent\",1]\n20:I[2992,[\"7586\",\"static/chunks/7586-203f1916d23f2407.js\",\"9079\",\"static/chunks/9079-00ef10f8b11f4d94.js\",\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"4400\",\"static/chunks/4400-4ff185dce52c1c17.js\",\"7531\",\"static/chunks/7531-64596f3089411931.js\",\"5061\",\"static/chunks/app/%5Blocale%5D/page-45ab204874fc5d47.js\"],\"PreloadCss\"]\n21:I[31982,[\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"1203\",\"static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js\"],\"SiteHeader\",1]\n22:I[87692,[\"1230\",\"static/chunks/1230-cd3b090f825a7d79.js\",\"1203\",\"static/chunks/app/%5Blocale%5D/layout-2406872f0cd5099b.js\"],\"default\"]\n17:T4dd,Adaptive intelligence is how AgentOS agents continuously improve their behavior without retraining. Five mechanisms work together: (1) Meta-reflective"])</script>
83<script>self.__next_f.push([1," prompt adaptation — the PromptBuilder assembles a different system prompt every turn, dynamically incorporating personality traits, mood state, conversation history, retrieved memories, and available tools. (2) Self-evaluating response quality — the self_evaluate tool scores the agent's own output and adjusts parameters like temperature, verbosity, and personality expression in real time. (3) Personality-modulated cognition — HEXACO traits shape how the agent processes information. High openness increases creative associations during memory retrieval; high conscientiousness strengthens retrieval-induced forgetting of irrelevant data. (4) Autonomous memory consolidation — the ConsolidationLoop prunes weak memories, strengthens frequently-accessed ones, and derives new insights from memory clusters, so the agent's knowledge improves over time without explicit training. (5) QueryRouter tiered classification — the system adapts retrieval depth based on query complexity. Simple questions get fast keyword lookup; complex questions trigger full hybrid RAG with deep research.18:T45a,Emergent behaviors are capabilities that agents develop at runtime rather than being explicitly programmed. AgentOS supports five forms: (1) Runtime tool forging — agents create new tools on the fly via forge_tool. The EmergentCapabilityEngine uses sandboxed JavaScript execution and LLM-as-judge evaluation to safely create, test, and promote tools. (2) Self-improving personality — agents adapt their HEXACO personality traits within bounded limits via adapt_personality. Mutations persist with Ebbinghaus decay — strong repeated adaptations stick, while weak ones fade naturally. (3) Dynamic skill management — agents enable or disable skills at runtime via manage_skills, adapting their behavioral repertoire to the task at hand. (4) Composable workflow creation — agents compose registered tools into multi-step pipelines via create_workflow, building new capabilities from existing building blocks. (5) Tiered tool promotion — "])</script>
83<script>self.__next_f.push([1,"forged tools progress through session, agent, and shared tiers. Tools that prove reliable (5+ successful uses, \u003e0.8 confidence score) auto-promote for cross-agent reuse.1a:T7f5,{\"@context\":\"https://schema.org\",\"@type\":\"SoftwareApplication\",\"name\":\"AgentOS\",\"alternateName\":[\"AgentOS Framework\",\"AgentOS Runtime\",\"AgentOS by Frame\"],\"description\":\"Open-source TypeScript AI agent framework. Build autonomous agents with cognitive memory, runtime tool generation, multi-agent orchestration, multimodal RAG, voice pipelines, and 11 LLM providers.\",\"url\":\"https://agentos.sh\",\"applicationCategory\":\"DeveloperApplication\",\"applicationSubCategory\":\"AI Agent Framework\",\"operatingSystem\":\"Any\",\"programmingLanguage\":\"TypeScript\",\"keywords\":\"TypeScript AI agent framework, AI agent runtime, open-source agent framework, multi-agent orchestration, cognitive memory, agent SDK, autonomous agents, AgentOS\",\"offers\":{\"@type\":\"Offer\",\"price\":\"0\",\"priceCurrency\":\"USD\",\"description\":\"Free and open-source (Apache 2.0)\"},\"author\":{\"@type\":\"Organization\",\"name\":\"Frame\",\"url\":\"https://frame.dev\"},\"downloadUrl\":\"https://www.npmjs.com/package/@framers/agentos\",\"installUrl\":\"https://www.npmjs.com/package/@framers/agentos\",\"softwareVersion\":\"0.1.x\",\"license\":\"https://opensource.org/licenses/Apache-2.0\",\"image\":\"https://agentos.sh/og-image-v2.png\",\"screenshot\":\"https://agentos.sh/og-image-v2.png\",\"featureList\":[\"TypeScript-first AI agent framework\",\"Multi-agent orchestration with 6 collaboration strategies\",\"Cognitive memory with Ebbinghaus decay and reconsolidation\",\"Runtime tool generation: agents write their own functions mid-task\",\"Specialist spawning: spawn_specialist mints new agents at runtime\",\"Multimodal RAG (text, image, audio, video)\",\"11 LLM providers (OpenAI, Anthropic, Google, Ollama, OpenRouter)\",\"37 channel adapters (Telegram, WhatsApp, Discord, Slack)\",\"Built-in safety: PII redaction, prompt injection defense, content moderation\",\"HEXACO personality system\",\"Voice pipeline (STT, TTS, VAD)\",\"80+ curated skills and extensions\",\"Self-h"])</script>
83<script>self.__next_f.push([1,"ostable, production-ready, Apache 2.0\"],\"sameAs\":[\"https://github.com/framerslab/agentos\",\"https://www.npmjs.com/package/@framers/agentos\",\"https://docs.agentos.sh\"]}"])</script>
83<script>self.__next_f.push([1,"b:[\"$\",\"$L16\",null,{\"formats\":\"$undefined\",\"locale\":\"en\",\"messages\":{\"metadata\":{\"title\":\"AgentOS (agentos.sh): Open Source TypeScript AI Agent Framework\",\"description\":\"AgentOS at agentos.sh is an open-source TypeScript AI agent framework. Build autonomous agents with cognitive memory, runtime tool generation, multi-agent orchestration, and 11 LLM providers. 85.6% on LongMemEval-S. Apache 2.0. By Frame.dev.\",\"ogTitle\":\"AgentOS (agentos.sh): Open Source TypeScript AI Agent Framework\",\"ogDescription\":\"AgentOS at agentos.sh: open-source TypeScript AI agent framework. Cognitive memory, runtime tool generation, multi-agent orchestration, voice pipelines, and 11 LLM providers. 85.6% on LongMemEval-S. Apache 2.0.\",\"twitterTitle\":\"AgentOS (agentos.sh): Open Source TypeScript AI Agent Framework\",\"twitterDescription\":\"AgentOS at agentos.sh: open-source TypeScript AI agent framework. Cognitive memory, runtime tool generation, multi-agent orchestration, and 11 LLM providers. 85.6% LongMemEval-S. Apache 2.0.\"},\"nav\":{\"features\":\"Features\",\"docs\":\"Docs\",\"guides\":\"Guides\",\"apiReference\":\"API Reference\",\"changelog\":\"Changelog\",\"openapi\":\"OpenAPI\",\"faq\":\"FAQ\",\"about\":\"About\",\"github\":\"GitHub\",\"marketplace\":\"Marketplace\"},\"hero\":{\"badge\":\"Open Source · TypeScript · Production-Ready\",\"title\":\"Open-Source TypeScript AI Agent Runtime\",\"subtitle\":\"Open-source TypeScript runtime for AI agents that remember, adapt, and spawn new tools. Agents write their own functions mid-task and the judge approves them; managers create specialists when the static roster does not cover the task. Cognitive memory grounded in neuroscience. 85.6% on LongMemEval-S. Apache 2.0.\",\"cycleWords\":[\"Adaptive\",\"Emergent\"],\"cycleWordsTail\":[\"emergent\",\"adaptive\"],\"headlinePairs\":[{\"top\":\"Production AI Agents\",\"bottom\":\"with Multimodal RAG\"},{\"top\":\"Multi-Agent Orchestration\",\"bottom\":\"across 37 Channels\"},{\"top\":\"Built-in Safety\",\"bottom\":\"PII Redaction \u0026 Injection Defense\"},{\"top\":\"11 LLM Providers\",\"bottom\":\"with 72 Curated Skills\"}],\"technicalHighlights\":[{\"title\":\"Streaming-first runtime\",\"detail\":\"Token-level delivery across 37 channel adapters and guardrails.\"},{\"title\":\"Multi-agent orchestration\",\"detail\":\"Parallel agents with auditable routing, approvals, and budgets.\"},{\"title\":\"Multimodal RAG\",\"detail\":\"7 vector backends with 4-tier auto-scaling (SQLite → HNSW → Postgres+pgvector → Qdrant), observational memory compression, 10 document loaders, and Ebbinghaus decay.\"},{\"title\":\"Self-hostable\",\"detail\":\"Deploy anywhere — your infrastructure, your data, your agents.\"}],\"visualStats\":[{\"label\":\"Agencies live\",\"value\":\"128\"},{\"label\":\"Approval SLA\",\"value\":\"3.2 min\"},{\"label\":\"Guardrail coverage\",\"value\":\"98%\"}],\"ctaPrimary\":\"Try Voice Chat Assistant\",\"ctaSecondary\":\"Star on GitHub\",\"marketplaceLead\":\"Need curated agents?\",\"marketplaceCta\":\"Marketplace coming soon\",\"codeLabel\":\"npm install @framers/agentos\",\"copyToast\":\"Copied to clipboard!\",\"copyAria\":\"Copy command\",\"getStarted\":\"Get Started\",\"viewOnGithub\":\"View on GitHub\",\"copiedToClipboard\":\"Copied to clipboard!\",\"stats\":{\"githubStars\":\"GitHub Stars\",\"contributors\":\"Contributors\",\"forks\":\"Forks\",\"securityScore\":\"Security Score\"},\"compliance\":{\"gdpr\":\"GDPR-ready guardrails\",\"soc2\":\"Self-hosted on your infra\",\"enterpriseSupport\":\"Enterprise Support Available\"},\"capabilities\":{\"ts\":{\"label\":\"TypeScript Native\",\"value\":\"Type-Safe\"},\"oss\":{\"label\":\"Open Source\",\"value\":\"Apache 2.0\"},\"streaming\":{\"label\":\"Real-time Streaming\",\"value\":\"Async Ready\"},\"memory\":{\"label\":\"Persistent Memory\",\"value\":\"Built-in\"}},\"heroVisualAlt\":\"AgentOS orchestration workspace preview\"},\"features\":{\"sectionTitle\":\"Developer-First Platform\",\"sectionSubtitle\":\"Production-grade infrastru
83cture for building, deploying, and scaling AI agents\",\"multiAgent\":{\"title\":\"Emergent Multi-Agent Coordination\",\"description\":\"Spawn researcher, analyst, critic, and executor personas from a single goal. Shared context, budgets, and approvals keep every branch in sync.\",\"pill\":\"🆕 v0.1.3\",\"bullet1\":\"Deterministic routing or adaptive fan-out\",\"bullet2\":\"Citations + memory stitched automatically\"},\"toolPacks\":{\"title\":\"Tool \u0026 Guardrail Packs\",\"description\":\"Ship opinionated extensions for search, vision, code execution, or custom APIs with built-in permissions and retries.\",\"pill\":\"Extension Ecosystem\",\"bullet1\":\"Manifest-driven tool registry\",\"bullet2\":\"Guardrail presets + compliance tags\"},\"language\":{\"title\":\"Language \u0026 Translation\",\"description\":\"Auto-detect locale and tone per surface. Personas stay fluent across voice, chat, and docs with translation providers of your choice.\",\"pill\":\"Multilingual Ready\",\"bullet1\":\"Persona-aware tone adaptation\",\"bullet2\":\"Fallback routing by locale\"},\"storage\":{\"title\":\"Storage \u0026 Deployment\",\"description\":\"7 vector store backends with 4-tier auto-scaling (SQLite → HNSW → Postgres+pgvector → Qdrant). One-command migration between all backends. Docker auto-setup for Qdrant and Postgres.\",\"pill\":\"Deploy Anywhere\",\"bullet1\":\"One-command vector backend migration\",\"bullet2\":\"Docker auto-setup for Qdrant + Postgres\"},\"workbench\":{\"title\":\"Local-First Workbench\",\"description\":\"Prototype entire agencies in-browser with persisted SQL, timeline scrubbers, and a built-in marketplace of personas.\",\"pill\":\"Offline Capable\",\"bullet1\":\"Inspect reasoning + OpenTelemetry traces live\",\"bullet2\":\"Sync artifacts back to the cloud\",\"bullet3\":\"Export full agencies as Markdown/JSON bundles\"},\"guardrails\":{\"title\":\"Guardrails \u0026 Safety\",\"description\":\"Built-in content moderation, PII detection, and compliance checks. Circuit breakers, cost guards, and stuck detection prevent runaway loops and excessive spending.\",\"pill\":\"Enterprise Safety\",\"bullet1\":\"Circuit breaker + cost guard + stuck detection\",\"bullet2\":\"Content filtering \u0026 approval workflows\"},\"streaming\":{\"title\":\"Streaming Runtime\",\"description\":\"Token-level streaming with real-time output. Built for low-latency interactions with support for SSE, WebSockets, and custom transports.\",\"pill\":\"Real-time\",\"bullet1\":\"Token-by-token delivery\",\"bullet2\":\"Multi-channel support\"},\"workflow\":{\"title\":\"Workflow Engine\",\"description\":\"Define complex multi-step workflows with conditional branching, parallel execution, and automatic retry logic.\",\"pill\":\"Orchestration\",\"bullet1\":\"Visual workflow builder\",\"bullet2\":\"Parallel task execution\"},\"capabilityDiscovery\":{\"title\":\"Smart Capability Discovery\",\"description\":\"Semantic, tiered discovery engine that reduces capability context by ~90%. Vector search + graph re-ranking finds the right tool, skill, or channel at the right time.\",\"pill\":\"~90% Token Reduction\",\"bullet1\":\"Three-tier token-budgeted context assembly\",\"bullet2\":\"Graph-based relationship re-ranking\"},\"memoryCompression\":{\"title\":\"Observational Memory Compression\",\"description\":\"LLM-based 3-10x memory compression with temporal reasoning. Self-improving consolidation that merges, prunes, and strengthens memories over time.\",\"pill\":\"3-10x Compression\",\"bullet1\":\"Temporal reasoning preserves causal chains\",\"bullet2\":\"Self-improving consolidation cycle\"},\"documentIngestion\":{\"title\":\"Document Ingestion Pipeline\",\"description\":\"10 document loaders with multimodal extraction. 3 PDF processing tiers (basic, OCR, vision). 4 chunking strategies (fixed, semantic, recursive, sliding window).\",\"pill\":\"10 Loaders\",\"bullet1\":\"3 PDF tiers: basic, OCR, vision\",\"bullet2\":\"4 chunking strategies including semantic\"},\"memoryPortability\":{\"title\":\"Memory Import / Export\",\"description\":\"Import and export agent memory in 10 formats including Obsidian vault, ChatGPT exports, Markdown, JSON, and CSV. One-command migration between all 7 vector backends.\",\"pill\":\"10 Formats\",\"bullet1\":\"Obsidian vault + ChatGPT import\",\"bullet2\":\"One-command backend migration\"}},\"marketplace\":{\"comingSoon\":\"Agent Marketplace Coming Soon\",\"description\":\"Share, buy, and sell your AgentOS agencies. We'll handle CI/CD, infrastru
83cture, payouts, and compliance so you can focus on adaptive, emergent, permanent intelligence.\"},\"caseStudies\":{\"comingSoon\":\"Case Studies Coming Soon\",\"description\":\"Real-world AgentOS implementations and production deployments\"},\"footer\":{\"backedBy\":\"Frame is a member of the \u003cdeepgramLink\u003eDeepgram Startup Program\u003c/deepgramLink\u003e and the \u003cnvidiaLink\u003eNVIDIA Inception Program\u003c/nvidiaLink\u003e\",\"tagline\":\"Open-source TypeScript AI agent runtime by Frame.dev\",\"resources\":\"Resources\",\"docsGuides\":\"Docs (Guides)\",\"apiReferenceTSDoc\":\"API Reference (TypeDoc/TSDoc)\",\"examples\":\"Examples\",\"releaseNotes\":\"Release Notes\",\"discord\":\"Discord\",\"company\":\"Company\",\"contact\":\"Contact\",\"legal\":\"Legal\",\"terms\":\"Terms\",\"privacy\":\"Privacy\",\"license\":\"Apache 2.0 (core) + MIT (agents, extensions, guardrails)\",\"copyright\":\"© {year} Frame.dev. All rights reserved.\"},\"cta\":{\"title\":\"Ready to Build Adaptive AI Agents?\",\"description\":\"Join developers building the next generation of intelligent systems with AgentOS.\",\"getStarted\":\"Get Started\",\"talkToUs\":\"Talk to Us\",\"badge\":\"Join the AI Revolution\",\"exploreLabel\":\"Or explore AgentOS today:\",\"tryVoiceChat\":\"Try Voice Chat\",\"browseCode\":\"Browse Code\",\"marketplace\":\"Marketplace\"},\"LanguageSwitcher\":{\"switchLanguage\":\"Switch language\",\"currentLanguage\":\"Current language: {language}\",\"selectLanguage\":\"Select language\"},\"skipToContent\":\"Skip to content\",\"skipToMain\":\"Skip to main content\",\"productCards\":{\"title\":\"Core Capabilities\",\"subtitle\":\"Everything you need to build production-grade AI agents with memory, personality, and autonomy\",\"stats\":{\"mechanisms\":\"Mechanisms\",\"personalityAxes\":\"Personality Axes\",\"decayCurve\":\"Decay Curve\",\"vectorBackends\":\"Vector Backends\",\"docLoaders\":\"Doc Loaders\",\"chunkStrategies\":\"Chunk Strategies\",\"strategies\":\"Strategies\",\"planningEngine\":\"Planning Engine\",\"workflowDsl\":\"Workflow DSL\",\"protocols\":\"Protocols\",\"guardrailMode\":\"Guardrail Mode\",\"connectionStability\":\"Connection Stability\",\"dimensions\":\"Dimensions\",\"presets\":\"Agent Presets\",\"styleAdaptation\":\"Style Adaptation\",\"securityTiers\":\"Security Tiers\",\"piiLayers\":\"PII Layers\",\"mlClassifiers\":\"ML Classifiers\",\"sttProviders\":\"STT Providers\",\"ttsProviders\":\"TTS Providers\",\"telephony\":\"Telephony\",\"providers\":\"LLM Providers\",\"fallbackChain\":\"Fallback Chain\",\"cliProviders\":\"CLI Providers\"},\"cards\":{\"cognitiveMemory\":{\"title\":\"Cognitive Memory\",\"description\":\"8 neuroscience-backed mechanisms with Ebbinghaus forgetting curve decay. Optional HEXACO personality vector biases retrieval; runs personality-neutral by default.\",\"features\":[\"Reconsolidation \u0026 RIF\",\"Involuntary recall \u0026 FOK\",\"Temporal gist extraction\",\"Source confidence decay\"],\"capabilities\":[\"Memory Reconsolidation\",\"Emotion Regulation\",\"Schema Encoding\"]},\"graphRag\":{\"title\":\"Graph-Based RAG\",\"description\":\"7 vector store backends with HyDE retrieval, GraphRAG community detection, and 10 document loaders\",\"features\":[\"HyDE hypothetical embeddings\",\"Louvain community detection\",\"4 chunking strategies\",\"SQLite to Neo4j backends\"],\"capabilities\":[\"HyDE Retrieval\",\"GraphRAG Pipeline\",\"Multi-Backend Storage\"]},\"orchestration\":{\"title\":\"Multi-Agent Orchestration\",\"description\":\"6 strategies from sequential to graph-based with Mission API, Tree-of-Thought planning, and workflow DSL\",\"features\":[\"Sequential \u0026 parallel\",\"Debate \u0026 review-loop\",\"Workflow DSL + checkpointing\",\"AgentGraph DAGs\"],\"capabilities\":[\"Mission API\",\"Workflow Engine\",\"Agent Graphs\"]},\"streaming\":{\"title\":\"Real-time Streaming\",\"description\":\"SSE and WebSocket support with in-flight content filtering, partial JSON output, and tool call progress events\",\"features\":[\"SSE + WebSocket\",\"In-flight guardrails\",\"Partial JSON streaming\",\"Tool progress events\"],\"capabilities\":[\"Event Streaming\",\"Streaming Guards\",\"Progress Tracking\"]},\"hexaco\":{\"title\":\"HEXACO Personality (Optional)\",\"description\":\"Opt-in 6-dimensional personality vector. When enabled, biases memory retrieval, communication style, and decision-making at the kernel level (not in the prompt). Omit it and the runtime stays personality-neutral.\",\"features\":[\"6-axis trait modeling\",\"Memory modulation\",\"Style adaptation\",\"9 agent presets\"],\"capabilities\":[\"Trait Engine\",\"Cognitive Modulation\",\"Adaptive Style\"]},\"guardrails\":{\"title\":\"Built-in Safety\",\"description\":\"Blocks prompt injection, redacts personal data, catches hallucinations, and scans generated code — configurable from permissive to locked-down\",\"features\":[\"4-tier PII redaction\",\"ONNX BERT classifiers\",\"Grounding guard (NLI)\",\"OWASP Top 10 scanning\"],\"capabilities\":[\"Security Tiers\",\"Content Classifiers\",\"Code Safety\"]},\"voice\":{\"title\":\"Voice Pipeline\",\"description\":\"12 STT and 12 TTS providers with real-time streaming, VAD, speaker diarization, and telephony support\",\"features\":[\"12 STT + 12 TTS\",\"VAD + diarization\",\"Real-time streaming\",\"Twilio/Telnyx/Plivo\"],\"capabilities\":[\"Speech Recognition\",\"Voice Activity Detection\",\"Telephony Integration\"]},\"providers\":{\"title\":\"11 LLM Providers\",\"description\":\"OpenAI, Anthropic, Gemini, Ollama, Groq, and 6 more with automatic fallback chains and provider-first API\",\"features\":[\"Auto fallback (402/429/5xx)\",\"Provider-first API\",\"CLI providers\",\"OpenRouter bridge\"],\"capabilities\":[\"Provider Registry\",\"Fallback Chains\",\"Model Resolution\"]}}}
83,\"enterprise\":{\"title\":\"Enterprise-Ready Infrastructure\",\"subtitle\":\"Safety guardrails, 11 LLM providers, 37 channels, and cognitive memory — production-grade from day one\",\"features\":{\"security\":{\"title\":\"Agent Safety\",\"items\":[\"Dangerous, Permissive, Balanced, Strict, Paranoid tiers\",\"LLM-based sentiment analysis with keyword fallback\",\"Configurable per-agent security presets\",\"Prompt injection defense built-in\"]},\"guardrails\":{\"title\":\"6 Guardrail Extensions\",\"items\":[\"PII redaction and anonymization\",\"ML classifier content moderation\",\"Code safety and sandboxed execution\",\"Grounding guard, topicality filter, content-policy rewriter\"]},\"compliance\":{\"title\":\"Privacy Architecture\",\"items\":[\"GDPR-aligned PII redaction (18 entity types)\",\"Data provenance and lineage tracking\",\"Cognitive memory with Ebbinghaus decay curves\",\"Source confidence decay for aging knowledge\"]},\"audit\":{\"title\":\"Audit \u0026 Observability\",\"items\":[\"OpenTelemetry-compatible telemetry (OTLP)\",\"Full audit logging for every agent action\",\"Real-time monitoring and anomaly detection\",\"Cost guards, circuit breakers, stuck detection\"]},\"soc2\":{\"title\":\"Audit-Friendly Architecture\",\"items\":[\"Per-event structured audit logs\",\"Cryptographic provenance chains\",\"Self-hosted deployment support\"]},\"scale\":{\"title\":\"Production Scale\",\"items\":[\"115 extensions, 88 curated skills\",\"11 LLM providers, 37 channel adapters\",\"Video, audio, and image generation\",\"Self-improving agents with tool forging\"]}},\"legend\":{\"fullyImplemented\":{\"title\":\"Fully Implemented\",\"description\":\"PII redaction, prompt injection defense, content moderation, audit logging, cognitive memory, and 115 extensions are production-ready today\"},\"inProgress\":{\"title\":\"Audit Trail\",\"description\":\"Per-redaction structured logs, provenance chains, and self-hosted data residency. Formal SOC 2 and GDPR attestations are the deploying organization's responsibility, not a library property.\"},\"enterpriseSupport\":{\"title\":\"Enterprise Support\",\"description\":\"Self-hosting, GPU orchestration, local model hosting, SSO, and dedicated support available\"}},\"cta\":{\"title\":\"AgentOS Enterprise\",\"description\":\"Self-hosting, advanced PII controls, OpenTelemetry telemetry, local model hosting, CI/CD pipelines, GPU orchestration, SSO, and dedicated support for production deployments.\",\"contact\":\"Contact [email protected]\"},\"trustBadge\":\"Trusted by industry leaders\",\"wipLabel\":\"WIP\",\"currentlyDevelopment\":\"Currently in development\"},\"multiAgent\":{\"title\":\"Multi-Agent Collaboration Network\",\"subtitle\":\"Researcher, Analyst, Creator, Critic, Executor stream insights in parallel\",\"description\":\"Orchestrator routes, memory persists - choose the right pattern for your use case\",\"patterns\":{\"consensus\":{\"title\":\"Consensus-Based Collaboration\",\"description\":\"Multiple agents vote on decisions with weighted confidence scores\",\"workflow\":{\"steps\":[\"All agents analyze input simultaneously\",\"Each agent proposes solution with confidence score\",\"Weighted voting based on expertise\",\"Consensus threshold validation\",\"Execute agreed-upon action\"],\"duration\":\"200-500ms\",\"throughput\":\"1000 decisions/min\"},\"useCases\":[{\"title\":\"Financial Trading Decisions\",\"description\":\"Multiple AI analysts vote on trade recommendations\",\"metrics\":[{\"label\":\"Accuracy\",\"value\":\"94%\"},{\"label\":\"Consensus Time\",\"value\":\"\u003c 200ms\"},{\"label\":\"Risk Reduction\",\"value\":\"67%\"}]},{\"title\":\"Content Moderation\",\"description\":\"Multiple agents evaluate content for policy violations\",\"metrics\":[{\"label\":\"False Positives\",\"value\":\"-82%\"},{\"label\":\"Coverage\",\"value\":\"99.9%\"},{\"label\":\"Review Speed\",\"value\":\"10k/hour\"}]},{\"title\":\"Medical Diagnosis Support\",\"description\":\"Specialist agents collaborate on patient diagnosis\",\"metrics\":[{\"label\":\"Diagnostic Accuracy\",\"value\":\"97%\"},{\"label\":\"Second Opinion Rate\",\"value\":\"100%\"},{\"label\":\"Time to Diagnosis\",\"value\":\"-45%\"}]}],\"proscons\":{\"pros\":[\"High accuracy through collective intelligence\",\"Built-in validation and error checking\",\"Reduces individual agent bias\",\"Transparent decision trail\"],\"cons\":[\"Higher latency than single agent\",\"Requires more compute resources\",\"Potential for deadlock scenarios\",\"Complex conflict resolution\"]}},\"sequential\":{\"title\":\"Sequential Pipeline\",\"description\":\"Agents process tasks in a defined order, each building on the previous\",\"
83workflow\":{\"steps\":[\"Researcher gathers initial data\",\"Analyst processes and structures information\",\"Creator generates solution\",\"Critic validates and refines\",\"Executor implements final output\"],\"duration\":\"500-2000ms\",\"throughput\":\"500 tasks/min\"},\"useCases\":[{\"title\":\"Document Processing Pipeline\",\"description\":\"Extract → Analyze → Summarize → Format → Deliver\",\"metrics\":[{\"label\":\"Accuracy\",\"value\":\"96%\"},{\"label\":\"Processing Time\",\"value\":\"1.2s/doc\"},{\"label\":\"Error Rate\",\"value\":\"\u003c 0.1%\"}]},{\"title\":\"Customer Support Escalation\",\"description\":\"Triage → Research → Solution → Review → Response\",\"metrics\":[{\"label\":\"Resolution Rate\",\"value\":\"89%\"},{\"label\":\"Avg Handle Time\",\"value\":\"45s\"},{\"label\":\"Customer Satisfaction\",\"value\":\"4.7/5\"}]},{\"title\":\"Code Review \u0026 Refactoring\",\"description\":\"Analyze → Identify Issues → Suggest Fixes → Test → Deploy\",\"metrics\":[{\"label\":\"Bug Detection\",\"value\":\"92%\"},{\"label\":\"Code Quality\",\"value\":\"+35%\"},{\"label\":\"Review Time\",\"value\":\"-60%\"}]}],\"proscons\":{\"pros\":[\"Clear workflow and accountability\",\"Easy to debug and monitor\",\"Predictable resource usage\",\"Simple error recovery\"],\"cons\":[\"Single point of failure risks\",\"Limited parallelization\",\"Higher total latency\",\"Bottleneck potential\"]}},\"parallel\":{\"title\":\"Parallel Execution\",\"description\":\"Agents work simultaneously on different aspects of the same task\",\"workflow\":{\"steps\":[\"Task decomposition by orchestrator\",\"Parallel assignment to all agents\",\"Concurrent processing\",\"Result aggregation\",\"Orchestrator merges outputs\"],\"duration\":\"100-300ms\",\"throughput\":\"5000 tasks/min\"},\"useCases\":[{\"title\":\"Real-time Market Analysis\",\"description\":\"Analyze stocks, bonds, forex, crypto simultaneously\",\"metrics\":[{\"label\":\"Coverage\",\"value\":\"10k assets\"},{\"label\":\"Update Frequency\",\"value\":\"100ms\"},{\"label\":\"Parallel Streams\",\"value\":\"128\"}]},{\"title\":\"Multi-language Translation\",\"description\":\"Translate content into multiple languages at once\",\"metrics\":[{\"label\":\"Languages\",\"value\":\"50+\"},{\"label\":\"Speed\",\"value\":\"\u003c 500ms\"},{\"label\":\"Consistency\",\"value\":\"98%\"}]},{\"title\":\"A/B Testing \u0026 Optimization\",\"description\":\"Test multiple variations simultaneously\",\"metrics\":[{\"label\":\"Variants\",\"value\":\"100+\"},{\"label\":\"Decision Time\",\"value\":\"\u003c 1s\"},{\"label\":\"Confidence\",\"value\":\"95%\"}]}],\"proscons\":{\"pros\":[\"Maximum throughput and speed\",\"Excellent scalability\",\"Fault isolation between agents\",\"Real-time processing capability\"],\"cons\":[\"Higher resource consumption\",\"Complex synchronization\",\"Potential race conditions\",\"Challenging debugging\"]}}},\"ui\":{\"workflowSteps\":\"Workflow Steps:\",\"latency\":\"Latency\",\"throughput\":\"Throughput\",\"pros\":\"Pros\",\"cons\":\"Cons\",\"useCases\":\"Use Cases\",\"case\":\"Case\",\"implementation\":\"Implementation\",\"showCode\":\"Show Code\",\"hideCode\":\"Hide Code\",\"setupTime\":\"Setup time: ~5 minutes\",\"performance\":\"Performance: Production-ready\",\"codePrompt\":\"Click \\\"Show Code\\\" to see how to implement {pattern} in your application.\"},\"agents\":{\"researcher\":\"Researcher\",\"analyst\":\"Analyst\",\"creator\":\"Creator\",\"critic\":\"Critic\",\"executor\":\"Executor\",\"orchestrator\":\"Orchestrator\"}},\"ecosystem\":{\"title\":\"The AgentOS Ecosystem\",\"subtitle\":\"Explore the open-source AgentOS ecosystem — multimodal RAG, multi-agent orchestration, 37 channel adapters, 11 LLM providers, and deployment kits for production AI teams.\",\"stats\":{\"repositories\":\"Repositories\",\"weeklyDownloads\":\"Weekly Downloads\",\"githubStars\":\"GitHub Stars\"},\"categories\":{\"core\":{\"title\":\"Core Frameworks\",\"description\":\"Runtime, skills content, and extension source code\"},\"infrastructure\":{\"title\":\"Infrastructure\",\"description\":\"Storage adapters, registries, and catalog SDKs\"},\"tools\":{\"title\":\"Tools\",\"description\":\"Developer tooling and debugging\"},\"apps\":{\"title\":\"Apps\",\"description\":\"Production applications and platforms\"},\"community\":{\"title\":\"Community\",\"description\":\"Forums, discussions, and resources\"}},\"starsSuffix\":\"stars\",\"updatedLabel\":\"Updated {date}\",\"cta\":{\"title\":\"Join the AgentOS Community\",\"description\":\"Contribute to open-source AI development, share your agents, and collaborate with developers worldwide.\",\"followOnGithub\":\"Follow on GitHub\",\"joinDiscord\":\"Join Discord Community\",\"contributeCode\":\"Contribute Code\"}},\"codeExamples\":{\"badge\":\"Code Examples\",\"title\":\"Build with AgentOS\",\"subtitle\":\"Production-ready code examples and patterns for building intelligent agent systems\",\"categories\":{\"all\":\"All Examples\",\"basic\":\"Basic\",\"advanced\":\"Advanced\",\"integration\":\"Integration\",\"deployment\":\"Deployment\"},\"selectExample\":\"Select Example\",\"tabs\":{\"synchronous\":\"Synchronous\",\"streaming\":\"Streaming\"},\"copyButton\":\"Copy code\",\"docsButton\":\"View Docs\",\"runButton\":\"Try in Playground\",\"examples\":{\"basicAgent\":{\"title\":\"Create Your First Agent\",\"description\":\"Basic agent setup with memory and tools\"},\"gmiRoles\":{\"title\":\"GMI Roles \u0026 Agencies\",\"description\":\"Implementing Generalised Mind Instance roles\"},\"memorySystem\":{\"title\":\"Personality \u0026 Cognitive Memory\",\"description\":\"HEXACO personality traits with Ebbinghaus decay and session history\"},\"toolIntegration\":{\"title\":\"RAG Pipeline\",\"description\":\"Ingest documents and retrieve context with HyDE embeddings\"},\"skillsIntegration\":{\"title\":\"Skills \u0026 Extensions\",\"description\":\"Browse the curated catalog and register extensions programmatically\"},\"realtimeStream\":{\"title\":\"Real-time Streaming Responses\",\"description\":\"Stream agent responses for better UX\"},\"deployment\":{\"title\":\"Production Deployment\",\"description\":\"Deploy AgentOS with Docker and Kubernetes\"},\"missionOrchestrator\":{\"title\":\"Mission Orchestrator\",\"description\":\"Describe a goal in natural language — the planner decomposes it into agents\"},\"voiceAgent\":{\"title\":\"Voice Agent with Tools\",\"description\":\"Build a voice-enabled agent with STT, TTS, and function calling\"},\"orchestrationGraph\":{\"title\":\"Execution Graph (Review Loop)\",\"description\":\"Build a cyclic review workflow with human approval gates\"},\"imageGeneration\":{\"title\":\"Structured Output (Zod)\",\"description\":\"Extract typed, validated data from unstructured text with any provider\"},\"emergentTools\":{\"title\":\"Emergent Tools\",\"description\":\"Enable runtime tool forging — agents create new tools when they hit capability gaps\"}}
83,\"cta\":{\"title\":\"Ready to Build Your First Agent?\",\"description\":\"Get started with AgentOS in less than 5 minutes. Install the SDK and follow our interactive tutorial.\",\"button\":\"Start Building\",\"time\":\"5 min\"}},\"socialProof\":{\"badge\":\"Social proof\",\"title\":\"Trusted deployments (revealing soon)\",\"subtitle\":\"We’re onboarding AI agency teams across finance, operations, and support. Logos and case studies unlock with their launches—stay tuned.\",\"reveal\":\"Reveal #{index}\",\"comingSoon\":\"Coming soon\",\"latestUpdates\":\"Latest updates\",\"cta\":{\"ask\":\"Want to be featured?\"},\"placeholders\":{\"helios\":\"Multi-market research pods (coming soon)\",\"northwind\":\"Product localization + CRM sync (coming soon)\",\"atlas\":\"Adaptive support routing (coming soon)\"}},\"gmiSection\":{\"badge\":\"Core Concept\",\"title\":\"Generalized Mind Instances\",\"subtitle\":\"Every agent in AgentOS is powered by a GMI — a persistent cognitive core with personality, memory, and autonomous decision-making. Not just a prompt wrapper: a mind.\",\"ctaExploreDocs\":\"Learn More\",\"codeExampleLabel\":\"Create a GMI\",\"comparisonTitle\":\"GMI vs Traditional Agents\",\"comparisonTraditionalLabel\":\"Traditional Agent\",\"comparisonGmiLabel\":\"AgentOS GMI\",\"layers\":{\"channels\":{\"name\":\"Channels \u0026 I/O\",\"description\":\"Telegram, Discord, Slack, WhatsApp, Twitter, and 32 more. Voice via 12 STT + 12 TTS providers. Every channel adapter streams tokens in real-time through a unified protocol.\"},\"guardrails\":{\"name\":\"Guardrails \u0026 Safety\",\"description\":\"Every output chunk is evaluated in real time. PII redaction, toxicity detection, hallucination grounding, and code safety scanning — configurable from permissive to fully locked down.\"},\"tools\":{\"name\":\"Tools \u0026 Extensions\",\"description\":\"Web search, browser automation, image generation, shell execution, social posting. 115 extensions and 88 curated skills. Capability discovery finds the right tool semantically.\"},\"orchestration\":{\"name\":\"Orchestration\",\"description\":\"Sequential, parallel, debate, hierarchical, review-loop, and graph strategies. Tree-of-Thought planning. Mission checkpointing with fork and resume semantics.\"},\"memory\":{\"name\":\"Memory\",\"description\":\"Ebbinghaus decay, reconsolidation, retrieval-induced forgetting, involuntary recall, feeling-of-knowing, temporal gist, schema encoding, source confidence decay. HyDE + GraphRAG retrieval across 7 vector backends.\"},\"personality\":{\"name\":\"Personality (HEXACO)\",\"description\":\"Honesty-Humility, Emotionality, Extraversion, Agreeableness, Conscientiousness, Openness. Traits modulate memory encoding, communication style, and risk tolerance.\"},\"llmCore\":{\"name\":\"LLM Core\",\"description\":\"11 LLM providers with automatic fallback chains. Provider-first API design. The thinking engine at the heart of every GMI.\"}},\"comparison\":{\"identity\":{\"traditional\":\"Stateless prompt\",\"gmi\":\"Persistent identity with HEXACO personality\"},\"memory\":{\"traditional\":\"No memory between sessions\",\"gmi\":\"Cognitive memory with Ebbinghaus decay + 8 mechanisms\"},\"behavior\":{\"traditional\":\"Fixed behavior\",\"gmi\":\"Style adaptation — learns your preferences\"},\"providers\":{\"traditional\":\"Single model\",\"gmi\":\"11 providers with automatic fallback\"},\"tools\":{\"traditional\":\"Manual tool wiring\",\"gmi\":\"Capability discovery — finds tools semantically\"},\"safety\":{\"traditional\":\"No safety\",\"gmi\":\"Real-time content safety with PII redaction and injection defense\"}}},\"faq\":{\"title\":\"Frequently Asked Questions\",\"subtitle\":\"Everything you need to know about AgentOS — from first install to production deployment.\",\"sections\":{\"general\":\"General\",\"technical\":\"Technical\",\"gettingStarted\":\"Getting Started\",\"enterprise\":\"Enterprise\",\"references\":\"Academic References\"},\"general\":{\"whatIsAgentOS\":{\"question\":\"What is AgentOS?\",\"answer\":\"AgentOS is an open-source TypeScript runtime for building production AI agents, developed by Frame.dev. It provides multi-agent orchestration, cognitive memory, multimodal RAG, built-in safety guardrails (PII redaction, prompt injection defense, content moderation), voice pipelines, 37 channel adapters, and 11 LLM provider integrations. AgentOS also powers the AI systems behind Wilds.ai, an AI-native gaming platform. Self-hostable and free.\"},\"isOpenSource\":{\"question\":\"Is AgentOS free and open source?\",\"answer\":\"Yes. The core runtime is licensed under Apache 2.0. Agent presets, extensions, and guardrails are MIT-licensed. You can use AgentOS commercially, modify it, and redistribute it under these license terms.\"},\"llmProviders\":{\"question\":\"What LLM providers are supported?\",\"answer\":\"AgentOS supports 11 LLM providers including OpenAI, Anthropic, Google Gemini, Groq, Together AI, Mistral, xAI (Grok), OpenRouter, Ollama, Claude Code CLI, Gemini CLI. Provider fallback chains allow automatic failover.\"},\"language\":{\"question\":\"What programming language is AgentOS built in?\",\"answer\":\"AgentOS is written entirely in TypeScript and runs on Node.js. The npm package is @framers/agentos. It works with any TypeScript or JavaScript project.\"},\"comparison\":{\"question\":\"How is AgentOS different from LangChain, CrewAI, or AutoGen?\",\"answer\":\"AgentOS is TypeScript-native (not Python), includes cognitive memory with Ebbinghaus decay curves and HEXACO personality modeling, offers 5 security tiers instead of basic prompt filtering, provides 37 built-in channel adapters (Discord, Telegram, Slack, email, and more), and supports 6 orchestration strategies including graph-based agent DAGs. It also ships with 88 curated skills and a capability discovery engine that reduces token usage by roughly 90%.\"},\"whatIsParacosm\":{\"question\":\"What is Paracosm?\",\"answer\":\"Paracosm (https://paracosm.agentos.sh) is a structured world-model simulation engine built on AgentOS. It compiles a JSON Sce
83narioPackage plus a HEXACO leader profile into a deterministic turn loop where AgentOS agents propose events, forge new TypeScript tools at runtime inside a hardened node:vm sandbox, and produce a counterfactual artifact you can fork and replay. Three modes share one schema: turn-loop civilization sims (Mars Genesis is the canonical example), batch-trajectory digital twins, and batch-point Monte-Carlo forecasts. Open-source under Apache 2.0, npm install paracosm. Live demo: https://paracosm.agentos.sh/sim. Source: https://github.com/framerslab/paracosm. Long-form launch post (why we built it, the wilds.ai vault story, the engine reference): https://agentos.sh/blog/paracosm-launch.\"},\"whatIsWildsAi\":{\"question\":\"What is Wilds.ai?\",\"answer\":\"Wilds.ai (https://wilds.ai) is an AI-native gaming platform built on AgentOS and Paracosm. AI characters with persistent cognitive memory and HEXACO personality traits inhabit shared world models and diverge from each other based on grounded LLM-as-judge events. Players can drop into existing worlds, fork them, or compile new ones in the browser. Wilds.ai is the canonical production deployment of the AgentOS + Paracosm stack and the official community / support channel (https://wilds.ai/discord) for both projects.\"},\"whatIsWunderland\":{\"question\":\"What is Wunderland?\",\"answer\":\"Wunderland (https://wunderland.sh) is a sister project under active development that packages a batteries-included CLI plus daemon on top of the AgentOS extension and skill registries. It ships a 28-command CLI (setup, chat, rag, agency, workflows, evaluate, provenance, knowledge, marketplace, and more), a 5-tier security model (dangerous / permissive / balanced / strict / paranoid), 8 agent presets, step-up HITL authorization, and the same cognitive memory plus graph-based RAG architecture documented in the AgentOS docs. Use @framers/agentos when you want a typescript SDK to embed in your application; use Wunderland when you want a global CLI you can install with pnpm add -g wunderland and configure with a wizard. Source at github.com/jddunn/wunderland (Apache 2.0 from 0.71.0). Preview release; APIs and CLI surface may still change. Getting-started guide: https://docs.agentos.sh/wunderland.\"}},\"technical\":{\"whatIsGMI\":{\"question\":\"What is a GMI?\",\"answer\":\"GMI stands for General Machine Intelligence — it is the cognitive core of each AgentOS agent. A GMI encapsulates the agent's personality traits (via HEXACO), memory systems, tool access, guardrail configuration, and communication style into a single persistent identity. GMIs maintain state across sessions and adapt their behavior based on accumulated experience.\"},\"hexaco\":{\"question\":\"What is HEXACO personality?\",\"answer\":\"HEXACO is a 6-factor personality model from psychology research (Ashton \u0026 Lee, 2004). The six dimensions are Honesty-Humility, Emotionality, Extraversion, Agreeableness, Conscientiousness, and Openness to Experience. In AgentOS, HEXACO traits modulate how agents communicate, remember, and make decisions. For example, high Openness increases memory reconsolidation during recall, while high Conscientiousness strengthens retrieval-induced forgetting of irrelevant associations.\"},\"cognitiveMemory\":{\"question\":\"How does cognitive memory work?\",\"answer\":\"AgentOS cognitive memory implements 8 mechanisms from cognitive science research: reconsolidation, retrieval-induced forgetting, involuntary recall, feeling-of-knowing, temporal gist extraction, schema encoding, source confidence decay, and emotion regulation. Memories decay over time following Ebbinghaus forgetting curves (Ebbinghaus, 1885), with spaced repetition and consolidation to preserve important information. Working memory follows Baddeley's multi-component model (Baddeley, 2000) with a volatile scratchpad for the current turn that consolidates upward into episodic, semantic, and observational memory tiers.\"},\"guardrails\":{\"question\":\"What guardrails are available?\",\"answer\":\"AgentOS provides 5 security tiers (dangerous, permissive, balanced, strict, paranoid) and 6 guardrail extension types: PII redaction (names, emails, phone numbers, SSNs), ML classifiers for prompt injection detection, code safety analysis, grounding verification against source documents, topicality enforcement to keep agents on-task, and content policy filtering for harmful or inappropriate output.\"},\"capabilityDiscovery\":{\"question\":\"What is the capability discovery system?\",\"answer\":\"The capability discovery engine replaces static tool and skill dumps with 3-tier semantic search, reducing token usage by roughly 90%. Tier 0 provides category summaries in about 150 tokens (always included). Tier 1 returns the top-5 semantic matches in about 200 tokens. Tier 2 delivers full schemas in about 1,500 tokens, loaded on demand. A meta-tool called discover_capabilities lets agents self-discover available tools at runtime.\"},\"queryRouter\":{\"question\":\"How does the QueryRouter work?\",\"answer\":\"The QueryRouter classifies incoming queries into 4 tiers: T0 (direct answer from context), T1 (single-tool invocation), T2 (multi-step plan requiring a planning engine), and T3 (multi-agent delegation requiring orchestration). This classification determines the execution strategy and resource allocation for each request.\"},\"adaptiveIntelligence\":{\"question\":\"What is adaptive intelligence?\",\"answer\":\"$17\"},\"emergentBehaviors\":{\"question\":\"What are emergent behaviors?\",\"answer\":\"$18\"},\"llmJudge\":{\"question\":\"Can I use an LLM as a judge instead of a human?\",\"answer\":\"Yes. AgentOS provides two complementary LLM-as-judge mechanisms. At the agency level, hitl.llmJudge() creates an approval handler that delegates decisions to an LLM. It evaluates each ApprovalRequest against configurable criteria and returns a structured decision with a confidence score. When confidence falls below the threshold, the decision escalates to a fallback handler (another hitl handler or a human). At the graph level, the humanNode() builder accepts a judge option that delegates the interrupt decision to an LLM. If the judge is confident, the graph continues without suspension. If confidence is low, execution falls through to a normal human interrupt. Both approaches support configurable model, provider, criteria, and confidence threshold. Example: hitl.llmJudge({ model: 'gpt-4o-mini', criteria: 'Is this response factually accurate?', confidenceThreshold: 0.8, fallback: hitl.cli() }).\"},\"guardrailOverride\":{\"question\":\"Can guardrails override HITL approvals?\",\"answer\":\"Yes. When hitl.guardrailOverride is enabled, guardrails run after an approval decision and can still block destructive or sensitive actions. This adds a second safety layer after auto-approve, human approval, or hitl.llmJudge(). The default post-approval guardrails include code-safety and pii-redaction, and 
83you can configure the list with hitl.postApprovalGuardrails. Disable the override only if you explicitly want the approval handler to have the final say.\"}},\"gettingStarted\":{\"install\":{\"question\":\"How do I install AgentOS?\",\"answer\":\"Install via npm: npm install @framers/agentos. AgentOS requires Node.js 18 or higher. For the CLI tools, install the companion package: npm install -g wunderland.\"},\"firstAgent\":{\"question\":\"How do I create my first agent?\",\"answer\":\"Import the agent function and configure it with a provider, instructions, and optional personality traits. Here is a minimal example:\\n\\nimport { agent } from '@framers/agentos'\\n\\nconst myAgent = agent({\\n  provider: 'openai',\\n  instructions: 'You are a helpful assistant.',\\n  personality: { openness: 0.8, conscientiousness: 0.9 },\\n  memory: { enabled: true },\\n})\\n\\nconst reply = await myAgent.send('Hello!')\\nconsole.log(reply.text)\"},\"voice\":{\"question\":\"How do I add voice capabilities?\",\"answer\":\"AgentOS supports multiple STT (speech-to-text) and TTS (text-to-speech) providers. Configure a voice pipeline by specifying STT and TTS providers in your agent configuration. Supported STT providers include Whisper (OpenAI), Deepgram, and AssemblyAI. TTS providers include ElevenLabs, OpenAI TTS, and Google Cloud TTS. Telephony integration is available for real-time voice interactions.\"},\"deploy\":{\"question\":\"How do I deploy AgentOS?\",\"answer\":\"AgentOS is self-hostable and runs anywhere Node.js runs. Deploy on your own infrastructure with Docker, on cloud providers like AWS, GCP, or Azure, or use platforms like Vercel, Railway, or Fly.io. The runtime uses SQLite by default for zero-config setup, with optional migration to Postgres, pgvector, Qdrant, or Neo4j for production scale.\"},\"offline\":{\"question\":\"Does AgentOS work offline?\",\"answer\":\"Yes. AgentOS works fully offline when paired with Ollama for local LLM inference. Ollama provides access to open-weight models like Llama, Mistral, and Phi locally. The SQLite storage backend and in-memory vector store require no external services. The CLI auto-detects Ollama and configures the agent to use it.\"}},\"enterprise\":{\"productionReady\":{\"question\":\"Is AgentOS production-ready?\",\"answer\":\"Yes. AgentOS lets you control exactly how much freedom your agents have — from wide open to fully locked down. It blocks prompt injection attacks, redacts personal data, moderates content in real time, requires approval for sensitive tool calls, enforces budget limits, and includes circuit breakers and rate limiting.\"},\"compliance\":{\"question\":\"How does AgentOS support GDPR and SOC 2 workflows?\",\"answer\":\"AgentOS ships PII redaction guardrails that strip personal data before it reaches LLM providers, structured audit logs for every redaction event, and self-hosted deployment so data stays on your infrastructure. The memory system supports deletion and export for data-subject access requests. GDPR and SOC 2 compliance are organizational responsibilities of the deploying team, not properties of the library itself.\"},\"support\":{\"question\":\"Is there enterprise support available?\",\"answer\":\"Yes. For production deployments, enterprise licensing, dedicated support, and custom integrations, contact the team at [email protected]. Enterprise support includes priority issue resolution, architecture consulting, and deployment assistance.\"},\"selfHosted\":{\"question\":\"Can I use AgentOS with my own infrastructure?\",\"answer\":\"Absolutely. AgentOS is designed to be self-hosted from the ground up. For self-hosted deployments, your data stays on your infrastructure with SQLite, Postgres, pgvector, Qdrant, or Neo4j. If you want a managed vendor-hosted vector database instead, Pinecone is also supported. Docker Compose configurations are provided for the self-hosted backends.\"}},\"references\":{\"title\":\"Academic References\",\"subtitle\":\"AgentOS cognitive memory and personality systems are grounded in peer-reviewed research.\",\"papers\":{\"ebbinghaus\":{\"label\":\"Ebbinghaus, H. (1885)\",\"title\":\"Memory: A Contribution to Experimental Psychology.\",\"detail\":\"The foundational study establishing the forgetting curve. AgentOS uses Ebbinghaus decay functions for automatic memory consolidation.\",\"url\":\"https://psychclassics.yorku.ca/Ebbinghaus/index.htm\"},\"baddeley\":{\"label\":\"Baddeley, A. D. (2000)\",\"title\":\"The episodic buffer: a new component of working memory?\",\"detail\":\"Introduced the 4-component model of working memory. AgentOS working memory is modeled on this architecture.\",\"url\":\"https://doi.org/10.1016/S1364-6613(00)01538-2\"},\"hexaco\":{\"label\":\"Ashton, M. C., \u0026 Lee, K. (2004)\",\"title\":\"Empirical, theoretical, and practical advantages of the HEXACO model of personality structure.\",\"detail\":\"The HEXACO model provides the 6-factor personality dimensions used in AgentOS agent identity.\",\"url\":\"https://doi.org/10.1207/s15327957pspr0802_1\"},\"louvain\":{\"label\":\"Blondel, V. D., Guillaume, J.-L., Lambiotte, R., \u0026 Lefebvre, E. (2008)\",\"t
83itle\":\"Fast unfolding of communities in large networks.\",\"detail\":\"The Louvain algorithm is used in AgentOS GraphRAG for community detection in knowledge graphs.\",\"url\":\"https://doi.org/10.1088/1742-5468/2008/10/P10008\"},\"hnsw\":{\"label\":\"Malkov, Y. A., \u0026 Yashunin, D. A. (2018)\",\"title\":\"Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs.\",\"detail\":\"HNSW is one of 7 vector store backends supported by AgentOS for efficient similarity search.\",\"url\":\"https://doi.org/10.1109/TPAMI.2018.2889473\"},\"reconsolidation\":{\"label\":\"Nader, K., Schafe, G. E., \u0026 LeDoux, J. E. (2000)\",\"title\":\"Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval.\",\"detail\":\"Basis for the memory reconsolidation mechanism — memories are rewritten each time they are recalled.\",\"url\":\"https://doi.org/10.1038/35021052\"},\"rif\":{\"label\":\"Anderson, M. C., Bjork, R. A., \u0026 Bjork, E. L. (1994)\",\"title\":\"Remembering can cause forgetting: Retrieval dynamics in long-term memory.\",\"detail\":\"Basis for retrieval-induced forgetting — retrieving one memory suppresses related competing memories.\",\"url\":\"https://doi.org/10.1037/0278-7393.20.5.1063\"},\"involuntaryRecall\":{\"label\":\"Berntsen, D. (2010)\",\"title\":\"The unbidden past: Involuntary autobiographical memories as a basic mode of remembering.\",\"detail\":\"Basis for involuntary recall — contextual cues trigger spontaneous memory activation.\",\"url\":\"https://doi.org/10.1177/0963721410370301\"}}},\"seeAlso\":\"See also\",\"links\":{\"vcaFaq\":\"VCA.Chat FAQ\",\"frameFaq\":\"Frame.dev FAQ\",\"frameDev\":\"Frame.dev\",\"lookingGlass\":\"The Looking Glass — AI newsletter \u0026 blog\"}},\"about\":{\"hero\":{\"weAreFramers\":\"we are the framers\",\"title\":\"About AgentOS\",\"description\":\"AgentOS is an open‑source TypeScript runtime for building adaptive, emergent AI agent systems. Built by \u003cframeLink\u003eFrame\u003c/frameLink\u003e, AgentOS powers intelligent applications with GMI orchestration, persistent memory, and enterprise‑grade guardrails. It also drives the AI systems behind \u003cwildsLink\u003eWilds.ai\u003c/wildsLink\u003e.\"},\"mission\":{\"title\":\"Our Mission\",\"p1\":\"We’re building the most advanced AI agent runtime for developers who want to create intelligent, adaptive systems without compromising on control, safety, or performance.\",\"p2\":\"AgentOS combines cutting-edge AI orchestration with production-ready infrastructure, enabling developers to build everything from simple chatbots to complex multi-agent systems with {span} memory architecture.\"},\"team\":{\"title\":\"The Team\",\"description\":\"AgentOS is developed by \u003cframeLink\u003eFrame\u003c/frameLink\u003e. We also build \u003cwildsLink\u003eWilds.ai\u003c/wildsLink\u003e, an AI-native gaming platform powered by AgentOS. We're a distributed team of engineers, designers, and AI researchers passionate about making advanced AI accessible to all developers.\",\"generalInquiries\":\"General Inquiries\",\"enterpriseSupport\":\"Enterprise \u0026 Support\",\"enterpriseSupportDesc\":\"For production deployments, enterprise licensing, and dedicated support\",\"hiring\":\"We are Hiring!\",\"joinUs\":\"Join us in building the future of autonomous agents.\"},\"connect\":{\"title\":\"Connect With Us\",\"github\":\"GitHub\",\"linkedin\":\"LinkedIn\",\"twitter\":\"Twitter\",\"frameDev\":\"Frame.dev\",\"ourCompany\":\"Our Company\"},\"license\":{\"title\":\"Open Source Licensing\",\"core\":\"AgentOS Core Engine\",\"coreDesc\":\"The core AgentOS runtime, GMI orchestration, and infrastructure\",\"extensions\":\"Extensions \u0026 Agents\",\"extensionsDesc\":\"Community-built agents, extensions, and guardrails\",\"footer\":\"AgentOS is open source under the Apache 2.0 license. For enterprise support and production deployments, contact us at \u003clink\[email protected]\u003c/link\u003e.\"},\"cta\":{\"title\":\"Ready to Build?\",\"viewOnGithub\":\"View on GitHub\",\"backToHome\":\"Back to Home\"}}
83,\"videoDemo\":{\"badge\":\"Live Demos\",\"title\":\"See AgentOS in Action\",\"subtitle\":\"Watch real workflows demonstrating agent creation, multi-agent collaboration, RAG memory, and more.\",\"comingSoon\":\"Demo video coming soon\",\"watchButton\":\"Watch Demo\",\"duration\":\"Duration\",\"videos\":{\"agentLoop\":{\"title\":\"Agent Loop\",\"description\":\"Tools, streaming, and traces in action\"},\"multiAgent\":{\"title\":\"Multi-Agent Coordination\",\"description\":\"Multiple agents working toward a shared goal\"},\"graphPipeline\":{\"title\":\"Graph Pipeline\",\"description\":\"5-node classify, lookup, draft, and approve flow\"},\"toolForge\":{\"title\":\"Tool Forge\",\"description\":\"Agent creates and uses its own tools\"}}},\"workbench\":{\"title\":\"AgentOS Workbench\",\"description\":\"The official desktop application for building, managing, and interacting with AgentOS experiences. Create powerful AI agents with a visual interface, test in real-time, and deploy anywhere.\",\"badge1\":\"MIT Licensed\",\"badge2\":\"Open Source\",\"feature1\":\"Visual agent builder with drag-and-drop workflows\",\"feature2\":\"Real-time testing and debugging tools\",\"feature3\":\"One-click deployment to cloud or local\",\"viewGithub\":\"View on GitHub\",\"downloadLatest\":\"Download Latest\",\"downloadComingSoon\":\"Download (coming soon)\"}
83,\"cognitive\":{\"badge\":\"Neuroscience-Backed\",\"title\":\"Cognitive Memory\",\"subtitle\":\"8 memory mechanisms inspired by cognitive science research. All modulated by the agent's HEXACO personality traits.\",\"mechanisms\":{\"reconsolidation\":\"Reconsolidation\",\"rif\":\"Retrieval-Induced Forgetting\",\"involuntaryRecall\":\"Involuntary Recall\",\"fok\":\"Feeling of Knowing\",\"temporalGist\":\"Temporal Gist\",\"schemaEncoding\":\"Schema Encoding\",\"sourceConfidence\":\"Source Confidence Decay\",\"emotionRegulation\":\"Emotion Regulation\"},\"mechanismDescriptions\":{\"reconsolidation\":\"Memories are rewritten each time they're recalled, updating with new context\",\"rif\":\"Retrieving one memory suppresses related competing memories\",\"involuntaryRecall\":\"Spontaneous memory activation triggered by contextual cues\",\"fok\":\"Metacognitive sense that an answer exists even when not immediately retrievable\",\"temporalGist\":\"Specific details fade while the gist and meaning is preserved over time\",\"schemaEncoding\":\"New information is encoded relative to existing knowledge structures\",\"sourceConfidence\":\"Certainty about where information came from decays faster than the information itself\",\"emotionRegulation\":\"Emotional valence of memories is modulated over time\"},\"modulators\":{\"reconsolidation\":\"Openness\",\"rif\":\"Conscientiousness\",\"involuntaryRecall\":\"Openness\",\"fok\":\"Extraversion\",\"temporalGist\":\"Conscientiousness\",\"schemaEncoding\":\"Openness\",\"sourceConfidence\":\"Honesty-Humility\",\"emotionRegulation\":\"Agreeableness\"},\"research\":{\"reconsolidation\":\"Each act of retrieval destabilizes the memory trace, allowing it to be updated with current context before re-stabilizing. Higher Openness amplifies the degree of rewrite, making memories more adaptive but less stable.\",\"rif\":\"Retrieving a target memory actively inhibits related competitors via lateral suppression. Higher Conscientiousness strengthens this suppression, keeping retrieval focused but increasing forgetting of peripheral associations.\",\"involuntaryRecall\":\"Contextual cues can trigger spontaneous activation of associated memories without deliberate search. Higher Openness lowers the activation threshold, producing more frequent involuntary recalls and serendipitous connections.\",\"fok\":\"A metacognitive judgment about the likelihood of successful retrieval, even before the answer surfaces. Higher Extraversion amplifies FOK signals, making the agent more confident in its sense of 'almost knowing.'\",\"temporalGist\":\"Over time, verbatim details decay while the semantic gist is preserved through abstraction. Lower Conscientiousness accelerates detail loss, leaving only the core meaning behind.\",\"schemaEncoding\":\"New information is assimilated into existing knowledge schemas, with schema-consistent material encoded more readily. Higher Openness allows more flexible schema modification, enabling novel information to reshape existing structures.\",\"sourceConfidence\":\"Memory for the source of information (who said it, where it was read) decays faster than the information content itself. Higher Honesty-Humility increases skepticism toward uncertain sources, treating low-confidence provenance as unreliable.\",\"emotionRegulation\":\"The emotional intensity attached to memories is actively modulated during retrieval and consolidation. Higher Agreeableness increases emotional smoothing, dampening extreme valence over time for more balanced recall.\"},\"citations\":{\"reconsolidation\":\"Nader, Schafe \u0026 LeDoux (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature, 406, 722-726.\",\"rif\":\"Anderson, Bjork \u0026 Bjork (1994). Remembering can cause forgetting: Retrieval dynamics in long-term memory. Journ
83al of Experimental Psychology: Learning, Memory, and Cognition, 20, 1063-1087.\",\"involuntaryRecall\":\"Berntsen (2010). The unbidden past: Involuntary autobiographical memories as a basic mode of remembering. Current Directions in Psychological Science, 19, 138-142.\",\"fok\":\"Koriat (1993). How do we know that we know? The accessibility model of the feeling of knowing. Psychological Review, 100, 609-639.\",\"temporalGist\":\"Reyna \u0026 Brainerd (1995). Fuzzy-trace theory: An interim synthesis. Learning and Individual Differences, 7, 1-75.\",\"schemaEncoding\":\"Bartlett (1932). Remembering: A Study in Experimental and Social Psychology. Cambridge University Press.\",\"sourceConfidence\":\"Johnson, Hashtroudi \u0026 Lindsay (1993). Source monitoring. Psychological Bulletin, 114, 3-28.\",\"emotionRegulation\":\"Gross (1998). The emerging field of emotion regulation: An integrative review. Review of General Psychology, 2, 271-299.\"},\"ebbinghausAnnotation\":\"Without rehearsal, 80% of information is lost within 24 hours. AgentOS cognitive memory applies spaced repetition and consolidation to flatten the curve.\",\"ctaLearnMore\":\"Explore Cognitive Memory\",\"ctaSoulWiki\":\"Markdown Memory Wiki\",\"memoryTypes\":{\"title\":\"4-Tier Memory Hierarchy\",\"subtitle\":\"From volatile scratchpads to permanent knowledge graphs — each tier consolidates upward automatically.\",\"working\":{\"name\":\"Working Memory\",\"description\":\"Short-term scratchpad for the current conversation turn. Cleared between sessions.\",\"detail\":\"~128k tokens, volatile\"},\"episodic\":{\"name\":\"Episodic Memory\",\"description\":\"Stores conversation episodes with temporal context. Auto-decays via Ebbinghaus curve.\",\"detail\":\"SQLite, Postgres, vector store\"},\"semantic\":{\"name\":\"Semantic Memory\",\"description\":\"Long-term knowledge extracted from episodes. GraphRAG with Louvain community detection.\",\"detail\":\"Vector embeddings + knowledge graph\"},\"observational\":{\"name\":\"Observational Memory\",\"description\":\"Compressed summaries of past interactions. 3-10x compression via LLM with temporal reasoning.\",\"detail\":\"Auto-consolidation from episodic\"},\"flowLabel\":\"Consolidation Pipeline\"},\"ragPipeline\":{\"title\":\"Multimodal RAG Pipeline\",\"subtitle\":\"Ingest documents, images, audio — retrieve with multiple strategies.\",\"steps\":{\"ingest\":{\"name\":\"Ingest\",\"description\":\"10 document loaders: PDF, DOCX, Markdown, HTML, CSV, JSON, TXT, images, audio, video\"},\"chunk\":{\"name\":\"Chunk\",\"description\":\"4 strategies: fixed-size, sentence, recursive, semantic\"},\"embed\":{\"name\":\"Embed\",\"description\":\"OpenAI, Ollama, Cohere, custom embeddings\"},\"store\":{\"name\":\"Store\",\"description\":\"7 vector store backends — click to explore\"},\"retrieve\":{\"name\":\"Retrieve\",\"description\":\"Multiple retrieval strategies for different use cases\"},\"rerank\":{\"name\":\"Rerank\",\"description\":\"Cross-encoder reranking for precision\"},\"context\":{\"name\":\"Context\",\"description\":\"Assembled into agent context with source attribution\"}},\"backends\":{\"sqlite\":{\"name\":\"SQLite\",\"description\":\"Zero-config, file-based. Great for development and single-agent deployments.\"},\"hnsw\":{\"name\":\"HNSW\",\"description\":\"In-process approximate nearest neighbor. Fast, no server needed.\"},\"pgvector\":{\"name\":\"Postgres+pgvector\",\"description\":\"Production-grade with pgvector extension. Scales to millions of vectors.\"},\"qdrant\":{\"name\":\"Qdrant\",\"description\":\"Dedicated vector DB. Best for high-throughput, multi-tenant deployments.\"},\"pinecone\":{\"name\":\"Pinecone\",\"description\":\"Managed cloud vector DB. Zero-ops, automatic scaling.\"},\"inmemory\":{\"name\":\"In-Memory\",\"description\":\"Fastest possible. For ephemeral or test workloads.\"},\"neo4j\":{\"name\":\"Neo4j\",\"description\":\"Graph-native. Best for GraphRAG with relationship-aware retrieval.\"}},\"strategies\":{\"semantic\":{\"name\":\"Semantic\",\"description\":\"Cosine similarity search over vector embeddings. The default for most use cases.\",\"badge\":\"default\"},\"hyde\":{\"name\":\"HyDE\",\"description\":\"Generates hypothetical answers to improve recall on abstract queries. Based on Gao et al. (2022).\",\"badge\":\"optional\",\"citation\":\"Precise Zero-Shot Dense Retrieval without Relevance Labels\",\"citationUrl\":\"https://arxiv.org/abs/2212.10496\"}
83,\"graphrag\":{\"name\":\"GraphRAG\",\"description\":\"Community-aware retrieval via knowledge graph with Louvain clustering. Based on Edge et al. (2024).\",\"badge\":\"\",\"citation\":\"From Local to Global: A Graph RAG Approach to Query-Focused Summarization\",\"citationUrl\":\"https://arxiv.org/abs/2404.16130\"},\"hybrid\":{\"name\":\"Hybrid\",\"description\":\"Keyword + semantic fusion using reciprocal rank. Best of both worlds.\",\"badge\":\"\"}}}},\"emergent\":{\"badge\":\"Runtime Intelligence\",\"title\":\"Emergent Capabilities\",\"subtitle\":\"Agents that forge their own tools and spawn specialist sub-agents at runtime\",\"statPills\":[\"Lifetime tiers: session / agent / shared\",\"Judge approval required\",\"Sandbox-executed\",\"Auto-promotion on reliability\"],\"tabs\":{\"toolForging\":\"Runtime Tool Forging\",\"agentSynthesis\":\"Agent Synthesis\",\"selfImproving\":\"Self-Improving Patterns\"},\"forging\":{\"step1\":\"Identify Gap\",\"step2\":\"Write Spec\",\"step3\":\"Judge Evaluates\",\"step4\":\"Tool Activated\",\"step5\":\"Agent Uses It\",\"description\":\"Runtime Tool Forging means an agent can write itself a new TypeScript function mid-decision, get it reviewed by a separate judge LLM, and execute it inside a hardened node:vm sandbox (5s default wall clock, 128 MB nominal memory budget, default-empty allowlist for fetch/fs/crypto, statically-banned eval/require/import/process/child_process), all before the next turn. Approved tools land in a discoverable index for reuse, so cost flattens after the first few forges in a session.\",\"details\":[\"Composable mode: chains existing tools via pipeline steps — safe by construction\",\"Sandbox mode: agent-written code runs in memory/time-bounded isolation\",\"Three-tier lifecycle: session-scoped, agent-scoped, or promoted to shared registry\",\"Dual-judge promotion panel: safety auditor + correctness reviewer must both approve\",\"Usage tracking with automatic tier promotion when reliability thresholds are met\"]},\"synthesis\":{\"orchestrator\":\"Orchestrator\",\"specialist1\":\"Data Analyst\",\"specialist2\":\"Code Reviewer\",\"specialist3\":\"Content Writer\",\"specialist4\":\"Fact Checker\",\"resultsMerge\":\"Results merge back to orchestrator\",\"description\":\"Agencies dynamically spawn new agents with specific personas and skills when the existing roster cannot cover a sub-task. The orchestrator identifies the capability gap, synthesises a specialist agent with tailored instructions, and merges its output back into the main workflow. Requires approval when HITL gates are configured.\"},\"loop\":{\"execute\":\"Execute\",\"observe\":\"Observe Outcome\",\"update\":\"Update Strategy\",\"executeBetter\":\"Execute Better\",\"circuitBreaker\":\"Circuit breaker on failures\",\"styleAdaptation\":\"Style adaptation (user prefs)\",\"memoryConsolidation\":\"Memory consolidation\",\"description\":\"Agents learn from failures and adapt their approach over time. Tool failure tracking triggers circuit breakers to avoid repeated errors. Style adaptation learns user preferences for formality, verbosity, and technicality. Memory consolidation strengthens useful memories and lets irrelevant ones decay via Ebbinghaus curves.\"},\"ctaExploreDocs\":\"Explore Emergent Capabilities\"},\"agencySection\":{\"badge\":\"Multi-Agent Orchestration\",\"title\":\"Parallel Agency\",\"subtitle\":\"Teams of specialized GMIs share execution state, cognitive memory, and RAG context across the roster. Six built-in strategies route outputs between brains: sequential, parallel, debate, review-loop, hierarchical, or graph DAG.\",\"codeExampleLabel\":\"Code Example\",\"sharedCapabilitiesTitle\":\"Shared Across All Strategies\",\"strategies\":{\"single-agent\":{\"name\":\"Single Agent\",\"description\":\"One GMI brain handling the full task on its own. No team coordination, no shared state. The baseline against which the strategies below show what an agency adds.\"},\"sequential\":{\"name\":\"Sequential\",\"description\":\"Agents run one after another in a linear pipeline. Each agent receives the previous agent's output as context, forming a progressive refinement chain.\"},\"parallel\":{\"name\":\"Parallel\",\"description\":\"All agents run concurrently. Their outputs are merged by a synthesis step that combines perspectives into a single coherent response.\"},\"debate\":{\"name\":\"Debate\",\"description\":\"Agents argue and refine a shared answer over multiple rounds. A judge evaluates arguments and selects the strongest position.\"},\"reviewLoop\":{\"name\":\"Review Loop\",\"description\":\"One agent produces output, another reviews and requests revisions. The loop continues until the reviewer approves or max rounds are reached.\"},\"review-loop\":{\"name\":\"Review Loop\",\"description\":\"One agent produces output, another reviews and requests revisions. The loop continues until the reviewer approves or max rounds are reached.\"},\"hierarchical\":{\"name\":\"Hierarchical\",\"description\":\"A coordinator agent dispatches sub-tasks to specialist agents via tool calls. The coordinator decides which agents to invoke and in what order at runtime.\"},\"graph\":{\"name\":\"Graph (DAG)\",\"description\":\"Agents declare explicit dependencies via depen
83dsOn. The orchestrator topologically sorts agents into tiers and runs each tier concurrently.\"}},\"diagrams\":{\"singleAgent\":\"Agent\",\"agentA\":\"Agent A\",\"agentB\":\"Agent B\",\"agentC\":\"Agent C\",\"input\":\"Input\",\"synthesis\":\"Synthesis\",\"proponent\":\"Proponent\",\"critic\":\"Critic\",\"judge\":\"Judge\",\"roundN\":\"Round N\",\"drafter\":\"Drafter\",\"reviewer\":\"Reviewer\",\"approved\":\"OK?\",\"output\":\"Out\",\"orchestrator\":\"Orchestrator\",\"specialistA\":\"Researcher\",\"specialistB\":\"Coder\",\"specialistC\":\"Writer\",\"researcher\":\"Researcher\",\"writer\":\"Writer\",\"illustrator\":\"Illustrator\"},\"sharedCapabilities\":{\"sharedMemory\":\"Shared Memory\",\"sharedRAG\":\"Shared RAG\",\"sharedExecState\":\"Shared Execution State\",\"hitlGates\":\"HITL Approval Gates\",\"streamingOutput\":\"Streaming Output\",\"resourceControls\":\"Resource Controls\",\"perAgentGuardrails\":\"Per-Agent Guardrails\"},\"sharedCapabilitiesScopeNoteLabel\":\"Scope:\",\"sharedCapabilitiesScopeNote\":\"Shared memory and roster apply within one \u003ccode\u003egenerate()\u003c/code\u003e call. Across \u003ccode\u003esession().send()\u003c/code\u003e turns only message history persists. See the \u003clink\u003eagency API docs\u003c/link\u003e for cross-turn memory patterns.\",\"ctaExploreDocs\":\"Explore Agency API\"},\"benchmarks\":{\"badge\":\"Memory benchmarks · LongMemEval · LOCOMO\",\"title\":\"Memory Benchmarks at Matched Reader\",\"subtitle\":\"Same gpt-4o reader, same dataset, same judge across every row. Cross-provider configurations are excluded because they cannot be reproduced from the methodology each vendor discloses publicly.\",\"tabS\":\"LongMemEval-S (115K tokens)\",\"tabM\":\"LongMemEval-M (1.5M tokens)\",\"highlightSAccuracy\":\"AgentOS LongMemEval-S\",\"highlightSValue\":\"85.6%\",\"highlightSCi\":\"0.4 points behind Emergence.ai SOTA (86%, closed-source SaaS) · +1.4 points above Mastra OM gpt-4o (84.23%) · $0.0090 per correct · 3.6-second median latency\",\"highlightMAccuracy\":\"AgentOS LongMemEval-M\",\"highlightMValue\":\"70.2%\",\"highlightMCi\":\"First open-source library above 65% on the 1.5M-token M variant · +4.5 points above Wu et al., ICLR 2025 baseline\",\"matrixIntro\":\"Each row is a full N=500 run at gpt-4o reader, with the same gpt-4o-2024-08-06 judge and the same rubric. Cross-provider configurations are excluded from this table.\",\"ctaReadAudit\":\"Read the Transparency Audit\",\"ctaSeeRunJsons\":\"See run JSONs on GitHub\",\"ctaBenchmarksPage\":\"Full Benchmarks Reference\",\"footnote\":\"Mastra OM 94.9%, Mem0 v3 93.4%, Hindsight 91.4%, agentmemory 96.2%, and Supermemory 85.2% all use cross-provider readers (gpt-5-mini, gemini-3-pro) or managed-platform infrastructure where the judge model, rubric, and seed are not published. The cross-vendor comparison above is what is independently reproducible from the methodology each vendor discloses.\"}}
83,\"now\":\"$undefined\",\"timeZone\":\"UTC\",\"children\":[\"$\",\"$L19\",null,{\"children\":[[\"$\",\"link\",null,{\"rel\":\"preconnect\",\"href\":\"https://api.github.com\",\"crossOrigin\":\"anonymous\"}],[\"$\",\"link\",null,{\"rel\":\"preconnect\",\"href\":\"https://img.shields.io\",\"crossOrigin\":\"anonymous\"}],[\"$\",\"link\",null,{\"rel\":\"dns-prefetch\",\"href\":\"https://api.npmjs.org\"}],[\"$\",\"link\",null,{\"rel\":\"dns-prefetch\",\"href\":\"https://static.cloudflareinsights.com\"}],[\"$\",\"link\",null,{\"rel\":\"dns-prefetch\",\"href\":\"https://www.googletagmanager.com\"}],[\"$\",\"link\",null,{\"rel\":\"dns-prefetch\",\"href\":\"https://www.google-analytics.com\"}],[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"$1a\"}}],[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"Organization\\\",\\\"name\\\":\\\"Frame\\\",\\\"alternateName\\\":[\\\"Frame.dev\\\",\\\"Frame team\\\"],\\\"url\\\":\\\"https://frame.dev\\\",\\\"logo\\\":\\\"https://agentos.sh/logo.png\\\",\\\"email\\\":\\\"mailto:[email protected]\\\",\\\"sameAs\\\":[\\\"https://github.com/framerslab\\\",\\\"https://twitter.com/framerslab\\\",\\\"https://www.linkedin.com/company/framerslab\\\",\\\"https://agentos.sh\\\",\\\"https://vca.chat\\\"]}\"}}],[\"$\",\"script\",null,{\"type\":\"application/ld+json\",\"dangerouslySetInnerHTML\":{\"__html\":\"{\\\"@context\\\":\\\"https://schema.org\\\",\\\"@type\\\":\\\"WebSite\\\",\\\"name\\\":\\\"AgentOS\\\",\\\"url\\\":\\\"https://agentos.sh\\\",\\\"potentialAction\\\":{\\\"@type\\\":\\\"SearchAction\\\",\\\"target\\\":\\\"https://docs.agentos.sh/search?q={search_term_string}\\\",\\\"query-input\\\":\\\"required name=search_term_string\\\"}}\"}}],[\"$\",\"$1b\",null,{\"fallback\":null,\"children\":[\"$\",\"$L1c\",null,{\"reason\":\"next/dynamic\",\"children\":[\"$\",\"$L1d\",null,{}]}]}],[\"$\",\"$1b\",null,{\"fallback\":null,\"children\":[\"$\",\"$L1c\",null,{\"reason\":\"next/dynamic\",\"children\":[\"$\",\"$L1e\",null,{}]}]}],[\"$\",\"$1b\",null,{\"fallback\":null,\"children\":[\"$\",\"$L1c\",null,{\"reason\":\"next/dynamic\",\"children\":[\"$\",\"$L1f\",null,{}]}]}],[\"$\",\"a\",null,{\"href\":\"#main-content\",\"className\":\"skip-to-content\",\"children\":\"Skip to content\"}],[\"$\",\"$1b\",null,{\"fallback\":[\"$\",\"header\",null,{\"className\":\"sticky top-0 z-50 w-full border-b border-[var(--color-border-subtle)] bg-[var(--color-background-primary)]/80 backdrop-blur-xl\",\"children\":[\"$\",\"div\",null,{\"className\":\"mx-auto flex h-16 max-w-7xl items-center justify-between px-4 sm:px-6 lg:px-8\",\"children\":[[\"$\",\"div\",null,{\"className\":\"flex items-center gap-2\",\"children\":[[\"$\",\"div\",null,{\"className\":\"h-8 w-8 rounded-lg bg-gradient-to-br from-violet-500 to-cyan-500 animate-pulse\"}],[\"$\",\"div\",null,{\"className\":\"h-5 w-24 rounded bg-[var(--color-background-secondary)] animate-pulse\"}]]}],[\"$\",\"div\",null,{\"className\":\"hidden lg:flex items-center gap-6\",\"children\":[[\"$\",\"div\",\"1\",{\"className\":\"h-4 w-16 rounded bg-[var(--color-background-secondary)] animate-pulse\"}],[\"$\",\"div\",\"2\",{\"className\":\"h-4 w-16 rounded bg-[var(--color-background-secondary)] animate-pulse\"}],[\"$\",\"div\",\"3\",{\"className\":\"h-4 w-16 rounded bg-[var(--color-background-secondary)] animate-pulse\"}],[\"$\",\"div\",\"4\",{\"className\":\"h-4 w-16 rounded bg-[var(--color-background-secondary)] animate-pulse\"}]]}],[\"$\",\"div\",null,{\"className\":\"flex items-center gap-2\",\"children\":[\"$\",\"div\",null,{\"className\":\"h-9 w-9 rounded-full bg-[var(--color-background-secondary)] animate-pulse\"}]}]]}]}],\"children\":[[\"$\",\"$L20\",null,{\"moduleIds\":[\"app/[locale]/layout.tsx -\u003e ../../components/site-header\"]}],[\"$\",\"$L21\",null,{}]]}],[\"$\",\"script\",null,{\"dangerouslySetInnerHTML\":{\"__html\":\"\\n              try {\\n                if (window.location.hash) {\\n                  var id = window.location.hash.slice(1);\\n                  var attempts = 0;\\n                  var poll = setInterval(function() {\\n                    var el = document.getElementById(id);\\n                    if (el) { el.scrollIntoView({ behavior: 'smooth', block: 'start' }); clearInterval(poll); }\\n                    if (++attempts \u003e= 20) clearInterval(poll);\\n                  }, 200);\\n                }\\n              } catch(e) {}\\n            \"}}],[\"$\",\"main\",null,{\"id\":\"main-content\",\"tabIndex\":-1,\"className\":\"focus:outline-none scroll-mt-24\",\"children\":[\"$\",\"$L8\",null,{\"parallelRouterKey\":\"children\",\"segmentPath\":[\"children\",\"$9\",\"children\"],\"error\":\"$undefined\",\"errorStyles\":\"$undefined\",\"errorScripts\":\"$undefined\",\"template\":[\"$\",\"$La\",null,{}],\"templateStyles\":\"$undefined\",\"templateScripts\":\"$undefined\",\"notFound\":\"$undefined\",\"notFoundStyles\":\"$undefined\"}]}],[\"$\",\"$L22\",null,{}],[\"$\",\"footer\",null,{\"className\":\"border-t border-purple-200/30 dark:border-purple-500/20 bg-gradient-to-br from-white/90 via-purple-50/30 to-pink-50/30 dark:from-black/80 dark:via-purple-950/40 dark:to-pink-950/40 backdrop-blur-lg py-12\",\"children\":[\"$\",\"div\",null,{\"className\":\"mx-auto w-full max-w-6xl px-6\",\"children\":[[\"$\",\"div\",null,{\"className\":\"grid md:grid-cols-4 gap-8 mb-8\",\"children\":[[\"$\",\"div\",null,{\"children\":[[\"$\",\"h3\",null,{\"className\":\"font-bold text-text-primary mb-3\",\"children\":\"AgentOS\"}],[\"$\",\"p\",null,{\"className\":\"text-sm text-text-secondary mb-3\",\"children\":\"Open-source TypeScript AI agent runtime by Frame.dev\"}],[\"$\",\"div\",null,{\"className\":\"flex gap-3\",\"children\":[[\"$\",\"a\",null,{\"href\":\"https://github.com/framerslab\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"aria-label\":\"FramersLab on GitHub\",\"children\":[[\"$\",\"svg\",null,{\"className\":\"w-5 h-5\",\"fill\":\"currentColor\",\"viewBox\":\"0 0 24 24\",\"children\":[\"$\",\"path\",null,{\"d\":\"M12 0c-6.626 0-12 5.373-12 12 0 5.302 3.438 9.8 8.207 11.387.599.111.793-.261.793-.577v-2.234c-3.338.726-4.033-1.416-4.033-1.416-.546-1.387-1.333-1.756-1.333-1.756-1.089-.745.083-.729.083-.729 1.205.084 1.839 1.237 1.839 1.237 1.07 1.834 2.807 1.304 3.492.997.107-.775.418-1.305.762-1.604-2.665-.305-5.467-1.334-5.467-5.931 0-1.311.469-2.381 1.236-3.221-.124-.303-.535-1.524.117-3.176 0 0 1.008-.322 3.301 1.23.957-.266 1.983-.399 3.003-.404 1.02.005 2.047.138 3.006.404 2.291-1.552 3.297-1.23 3.297-1.23.653 1.653.242 2.874.118 3.176.77.84 1.235 1.911 1.235 3.221 0 4.609-2.807 5.624-5.479 5.921.43.372.823 1.102.823 2.222v3.293c0 .319.192.694.801.576 4.765-1.589 8.199-6.086 8.199-11.386 0-6.627-5.373-12-12-12z\"}]}],[\"$\",\"span\",null,{\"className\":\"sr-only\",\"children\":\"GitHub\"}]]}],[\"$\",\"a\",null,{\"href\":\"https://www.linkedin.com/company/framerslab\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"aria-label\":\"Frame on LinkedIn\",\"children\":[[\"$\",\"svg\",null,{\"className\":\"w-5 h-5\",\"fill\":\"currentColor\",\"viewBox\":\"0 0 24 24\",\"children\":[\"$\",\"path\",null,{\"d\":\"M20.447 20.452h-3.554v-5.569c0-1.328-.027-3.037-1.852-3.037-1.853 0-2.136 1.445-2.136 2.939v5.667H9.351V9h3.414v1.561h.046c.477-.9 1.637-1.85 3.37-1.85 3.601 0 4.267 2.37 4.267 5.455v6.286zM5.337 7.433c-1.144 0-2.063-.926-2.063-2.065 0-1.138.92-2.063 2.063-2.063 1.14 0 2.064.925 2.064 2.063 0 1.139-.925 2.065-2.064 2.065zm1.782 13.019H3.555V9h3.564v11.452zM22.225 0H1.771C.792 0 0 .774 0 1.729v20.542C0 23.227.792 24 1.771 24h20.451C23.2 24 24 23.227 24 22.271V1.729C24 .774 23.2 0 22.222 0h.003z\"}]}],[\"$\",\"span\",null,{\"className\":\"sr-only\",\"children\":\"LinkedIn\"}]]}],[\"$\",\"a\",null,{\"href\":\"https://twitter.com/framerslab\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"aria-label\":\"FramersLab on Twitter\",\"children\":[[\"$\",\"svg\",null,{\"className\":\"w-5 h-5\",\"fill\":\"currentColor\",\"viewBox\":\"0 0 24 24\",\"children\":[\"$\",\"path\",null,{\"d\":\"M23.953 4.57a10 10 0 01-2.825.775 4.958 4.958 0 002.163-2.723c-.951.555-2.005.959-3.127 1.184a4.92 4.92 0 00-8.384 4.482C7.69 8.095 4.067 6.13 1.64 3.162a4.822 4.822 0 00-.666 2.475c0 1.71.87 3.213 2.188 4.096a4.904 4.904 0 01-2.228-.616v.06a4.923 4.923 0 003.946 4.827 4.996 4.996 0 01-2.212.085 4.936 4.936 0 004.604 3.417 9.867 9.867 0 01-6.102 2.105c-.39 0-.779-.023-1.17-.067a13.995 13.995 0 007.557 2.209c9.053 0 13.998-7.496 13.998-13.985 0-.21 0-.42-.015-.63A9.935 9.935 0 0024 4.59z\"}]}],[\"$\",\"span\",null,{\"className\":\"sr-only\",\"children\":\"Twitter\"}]]}]]}]]}],[\"$\",\"div\",null,{\"children\":[[\"$\",\"h3\",null,{\"className\":\"font-bold text-text-primary mb-3\",\"children\":\"Resources\"}],[\"$\",\"ul\",null,{\"className\":\"space-y-2 text-sm\",\"children\":[[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://docs.agentos.sh\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Docs (Guides)\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://docs.agentos.sh/api\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"API Reference (TypeDoc/TSDoc)\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/guides\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Guides\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/#code\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Examples\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://github.com/framerslab/agentos/releases\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Release Notes\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://github.com/framerslab/agentos\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"GitHub\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://wilds.ai/discord\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Discord\"}]}]]}]]}],[\"$\",\"div\",null,{\"children\":[[\"$\",\"h3\",null,{\"className\":\"font-bold text-text-primary mb-3\",\"children\":\"Company\"}],[\"$\",\"ul\",null,{\"className\":\"space-y-2 text-sm\",\"children\":[[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://frame.dev\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Frame.dev\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://wilds.ai\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Wilds.ai\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://paracosm.agentos.sh\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Paracosm\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"https://wunderland.sh\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Wunderland\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/about\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"About\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/faq\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"FAQ\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/contact\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Contact\"}]}]]}]]}],[\"$\",\"div\",null,{\"children\":[[\"$\",\"h3\",null,{\"className\":\"font-bold text-text-primary mb-3\",\"children\":\"Legal\"}],[\"$\",\"ul\",null,{\"className\":\"space-y-2 text-sm\",\"children\":[[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/legal/terms\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Terms\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/legal/privacy\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Privacy\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/legal/security\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Security\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"a\",null,{\"href\":\"/en/legal/cookies\",\"className\":\"text-text-secondary hover:text-accent-primary transition-colors\",\"children\":\"Cookies\"}]}],[\"$\",\"li\",null,{\"children\":[\"$\",\"span\",null,{\"className\":\"text-text-secondary\",\"children\":\"Apache 2.0 (core) + MIT (agents, extensions, guardrails)\"}]}]]}]]}]]}],[\"$\",\"div\",null,{\"className\":\"pt-6 border-t border-purple-200/30 dark:border-purple-500/20 text-center\",\"children\":[[\"$\",\"p\",null,{\"className\":\"text-text-secondary\",\"children\":\"© 2026 Frame.dev. All rights reserved.\"}],[\"$\",\"p\",null,{\"className\":\"text-text-secondary text-sm mt-2\",\"children\":[\"Frame is a member of the \",[\"$\",\"a\",\"deepgramLink0\",{\"href\":\"https://deepgram.com/startups\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-accent-primary hover:underline\",\"children\":[\"Deepgram Startup Program\"]}],\" and the \",[\"$\",\"a\",\"nvidiaLink0\",{\"href\":\"https://www.nvidia.com/en-us
83/startups/\",\"target\":\"_blank\",\"rel\":\"noopener noreferrer\",\"className\":\"text-accent-primary hover:underline\",\"children\":[\"NVIDIA Inception Program\"]}]]}]]}]]}]}]]}]}]\n"])</script>
83<script type="module" src="https://static.cloudflareinsights.com/beacon.min.js/v31edd6df95cf4e85bb4c19e7a9bdbcba1788362987495" integrity="sha512-iIg7k2xntmwu6/uSb5tpc/hySgZc4eoL31yB29W6tJFo2akwjPWcEqnCEdJvGexCL0KEQwVYv5BlowfhVz26hg==" data-cf-beacon='{"version":"2024.11.0","token":"614d61ebed394cd5ba863ddb7e567c41","r":1,"spa":2}' crossorigin="anonymous"></script>
83
84</body></html>

Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.