1import{u as s,j as e,L as t,F as a,ai as i}from"./index-2XUgqg3-.js";import{S as r}from"./SeoArticleShell-BG9JKqQM.js";import{M as n}from"./MeetKonshusCta-CNcsBgP5.js";import"./PatentNotice-DfPt3Npp.js";import"./StatusFooterIndicator-CAPBelom.js";import"./status.functions-DgF9zIeD.js";import"./shield-check-DNkOXpTg.js";import"./arrow-right-03-YyPmt.js";function u(){return s("chatgpt_memory_mcp_view"),e.jsxs(r,{children:[e.jsxs("article",{className:"mx-auto max-w-3xl px-6 pt-12 pb-4",children:[e.jsx("p",{className:"text-xs uppercase tracking-wider text-ink/50 mb-3",children:"A field note · ~6 min read"}),e.jsx("h1",{className:"font-serif text-4xl md:text-5xl leading-tight text-ink mb-5",children:"ChatGPT memory over MCP"}),e.jsx("p",{className:"text-lg text-ink/75 leading-relaxed",children:"Every complaint about ChatGPT's memory comes down to the same constraint: it has to fit inside the prompt of every conversation, so it has to be tiny. MCP sidesteps the constraint entirely. The model stops carrying you around and starts looking you up."})]}),e.jsxs("section",{className:"mx-auto max-w-3xl px-6 py-8 space-y-4 text-ink/80 leading-relaxed",children:[e.jsx("h2",{className:"font-serif text-2xl md:text-3xl text-ink !mt-4",children:"Carrying versus querying"}),e.jsxs("p",{children:["Built-in memory is ",e.jsx("em",{children:"carried"}),". Every saved fact is injected into every conversation whether it is relevant or not, which is exactly why it has to stay around 1,200â2,000 tokens. Multiply a larger bucket by hundreds of millions of users and the arithmetic stops working."]}),e.jsxs("p",{children:["An MCP memory server is ",e.jsx("em",{children:"queried"}),". Nothing is injected up front. When the model needs to know how you like documents structured, it calls a tool and asks. Cost scales with relevance rather than with the size of your history, which is why the ceiling effectively disappears."]})]}),e.jsxs("section",{className:"mx-auto max-w-3xl px-6 py-8 space-y-4 text-ink/80 leading-relaxed",children:[e.jsx("h2",{className:"font-serif text-2xl md:text-3xl text-ink",children:"What that changes in practice"}),e.jsx("div",{className:"overflow-x-auto rounded-xl border border-border/60",children:e.jsxs("table",{className:"w-full text-sm",children:[e.jsx("thead",{className:"bg-muted/40 text-ink",children:e.jsxs("tr",{children:[e.jsx("th",{className:"text-left p-3 font-medium",children:" "}),e.jsx("th",{className:"text-left p-3 font-medium",children:"Built-in memory"}),e.jsx("th",{className:"text-left p-3 font-medium",children:"MCP memory server"})]})}),e.jsxs("tbody",{className:"divide-y divide-border/50",children:[e.jsxs("tr",{children:[e.jsx("td",{className:"p-3 text-ink",children:"Size"}),e.jsx("td",{className:"p-3",children:"~8â12 short facts"}),e.jsx("td",{className:"p-3",children:"No practical ceiling"})]}),e.jsxs("tr",{children:[e.jsx("td",{className:"p-3 text-ink",children:"Setup"}),e.jsx("td",{className:"p-3",children:"None"}),e.jsx("td",{className:"p-3",children:"Developer mode plus a URL"})]}),e.jsxs("tr",{children:[e.jsx("td",{className:"p-3 text-ink",children:"Exportable"}),e.jsx("td",{className:"p-3",children:"No"}),e.jsx("td",{className:"p-3",children:"Yes â it is your store"})]}),e.jsxs("tr",{children:[e.jsx("td",{className:"p-3 text-ink",children:"Works in Claude too"}),e.jsx("td",{className:"p-3",children:"No"}),e.jsx("td",{className:"p-3",children:"Yes, same server"})]}),e.jsxs("tr",{children:[e.jsx("td",{className:"p-3 text-ink",children:"Survives a model change"}),e.jsx("td",{className:"p-3",children:"Not guaranteed"}),e.jsx("td",{className:"p-3",children:"Yes"})]}),e.jsxs("tr",{children:[e.jsx("td",{className:"p-3 text-ink",children:"Works on mobile"}),e.jsx("td",{className:"p-3",children:"Yes"}),e.jsx("td",{className:"p-3",children:"No â web only"})]})]})]})}),e.jsx("p",{children:"That last row is a real limitation, not a footnote. Developer mode is web-only, so an MCP-backed memory does not follow you to the phone. For most people the honest answer is to use both: built-in memory for the handful of facts that should always be present, MCP for the depth."})]}),e.jsx("section",{className:"mx-auto max-w-3xl px-6 py-6",children:e.jsxs("div",{className:"rounded-xl border border-border/60 bg-muted/30 p-5",children:[e.jsx("p",{className:"text-xs uppercase tracking-wider text-ink/50 mb-2",children:"Short answer"}),e.jsx("p",{className:"text-ink leading-relaxed",children:"MCP does not make ChatGPT's memory bigger. It makes the size of ChatGPT's memory stop mattering, by putting the real store outside and letting the model ask. And because it is outside, the same memory answers Claude tomorrow."})]})}),e.jsxs("section",{className:"mx-auto max-w-3xl px-6 py-8 space-y-4 text-ink/80 leading-relaxed",children:[e.jsx("h2",{className:"font-serif text-2xl md:text-3xl text-ink",children:"Getting there"}),e.jsxs("p",{children:["Enable Developer mode, add"," ",e.jsx("code",{className:"text-sm text-ink",children:"https://konshus.ai/api/mcp"}),", sign in. Full walkthrough in"," ",e.jsx(t,{to:"/how-to-add-mcp-connector-to-chatgpt",className:"text-gold underline underline-offset-2",children:"how to add an MCP connector to ChatGPT"}),"."]}),e.jsxs("p",{children:["A vault only helps if there is something in it, so the useful first move is usually importing your existing history rather than starting from a blank page â see"," ",e.jsx(t,{to:"/backup-chatgpt",className:"text-gold underline underline-offset-2",children:"how to back up ChatGPT"}),"."]})]}),e.jsx("section",{className:"mx-auto max-w-3xl px-6 py-8 space-y-4 text-ink/80 leading-relaxed",children:e.jsxs("p",{children:["Related reading:"," ",e.jsx(t,{to:"/chatgpt-memory-limits",className:"text-gold underline underline-offset-2",children:"how much ChatGPT actually remembers"}),","," ",e.jsx(t,{to:"/chatgpt-memory-full",className:"text-gold underline underline-offset-2",children:"when memory is full"}),", and"," ",e.jsx(t,{to:"/chatgpt-mcp",className:"text-gold underline underline-offset-2",children:"ChatGPT MCP explained"}),"."]})}),e.jsx(a,{items:i}),e.jsx(n,{variant:"bottom",headline:"Eight slots is not a memory.",body:"Konshus holds years of you outside any provider and answers ChatGPT over MCP â so the model can look you up instead of guessing."})]})}
1export{u as component};
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.