One place for your context. Every tool reads it.

Capture context anywhere, manage it once, and reach it from every AI tool you use. 29 MCP tools, 75+ file formats, Deep Search inside documents.

Works with the MCP clients you already use

  • Cursor
  • Claude
  • ChatGPT
  • VS Code
  • Windsurf
  • Amp
  • Factory
  • Grok

Developer note

I stopped losing research to closed tabs. The extension saves the page or the chat as a Markdown document in my repository, and Claude Code can read it the next morning without me pasting anything.

Everything the repository does for you.

  • 75+ file formats

    PDF, Word, Excel, PowerPoint, images with OCR, and code files, parsed with LlamaIndex into searchable context.

  • Three capture paths

    Save visible ChatGPT and Claude chats as role-tagged Markdown, capture a public page URL, or send selected page content.

  • Collections

    Group prompts and documents into named folders scoped per project, with tags and explicit membership.

  • Full version history

    Prompt content edits and document changes are versioned. Restore an earlier version, and pin the version a shared link shows.

  • MCP server, 29 tools

    Plug your collections into ChatGPT, Claude, Cursor, Codex, Claude Code, and other MCP clients.

  • Deep Search

    Hierarchical vector search that lands on the exact passage inside a document, then expands for context.

How context moves.

Read the quickstart
Diagram of how context moves through Context Repo. People save context from the dashboard, the Chrome extension, and file uploads into one repository of versioned prompts, indexed documents, and collections, with Search, Deep Search, and Reason. AI chats (ChatGPT, Claude, Grok), coding agents (Claude Code, Codex, Droid), and code editors (Cursor, VS Code, Windsurf) read from it and write to it over MCP, and your own code does the same over the REST API. A highlighted path follows a chat saved with the Chrome extension into Documents and out to a coding agent.
  • You save it. From the Chrome extension, a file upload in any of 75+ formats, or the dashboard editor.

  • Context Repo keeps it. Version history on every prompt and document, plus indexing so Deep Search finds the exact passage.

  • Agents use it and add to it. ChatGPT, Claude, Cursor, and Claude Code read and save through 29 MCP tools. Your apps use the REST API.

Three surfaces. One repository.

The problem

Every AI chat keeps your history to itself. The plan you worked out in ChatGPT never reaches Claude or Grok, so you paste it back in by hand.

How Context Repo helps

Save what matters from ChatGPT, Claude, or Grok, and pull it into any chat that connects through MCP or a plugin. Your context goes where you go.

A real ChatGPT conversation in dark mode with the sidebar collapsed to its icon rail and the ChatGPT logo: the user's message tags the Context Repo plugin and asks it to deep search "restore orders-db", and ChatGPT's expanded tool card shows Context Repo, Deep search, and the request {query: "restore orders-db"}.

Developer note

One MCP server for every tool I use. Cursor, Claude, and ChatGPT read the same prompts and specs, so the context is the same in whichever one I have open.

Built for every workflow. Researching, coding, or writing.

  • Diagram of the research workflow. In your tools, the Chrome extension saves chats, pages, and selections into Context Repo, which holds indexed documents, versioned prompts, and organized collections, with Search, Deep Search, and Expand. AI tools (ChatGPT, Claude, Cursor) retrieve from it over MCP or the API. The highlighted path runs from the Chrome extension into Documents and out to the AI tools.

    Researchers and knowledge workers

    • Capture public page URLs or selected content from Chrome
    • Semantic catalog search, plus Deep Search inside documents
    • Bring grounded context into ChatGPT, Claude, or Cursor over MCP

    Capture papers, articles, and documentation. Build a knowledge base your connected AI tools can retrieve from once it is indexed.

  • Diagram of the developer workflow. In your tools, MCP clients (Claude, Cursor, VS Code) connect through the npx bridge with an API key and write prompts into Context Repo, which holds indexed documents, versioned prompts, and organized collections, with Search, Deep Search, and Expand. Your code reads them over the REST API with an API key. The highlighted path runs from the MCP clients into Prompts and out to your code.

    Developers and engineers

    • 29 MCP tools for Claude, Cursor, and VS Code
    • Version-controlled prompts
    • API keys for programmatic access

    Store coding prompts, API docs, and specs. Edit prompts from Cursor or your IDE, and your code reads the new version with an API key.

  • Diagram of the writing workflow. In your tools, the dashboard holds your prompt templates and style guides and groups them into a collection in Context Repo, which holds indexed documents, versioned prompts, and organized collections, with Search, Deep Search, and Expand. AI chats (ChatGPT, Claude, Grok) read the collection over MCP, signed in with OAuth or an API key. The highlighted path runs from the dashboard into Collections and out to the AI chats.

    Content creators and writers

    • Prompt templates with variables
    • Collections with tags and explicit membership
    • MCP server for Claude, ChatGPT, and more

    Keep writing prompts, style guides, and references together. Hold one voice across everything you publish.

  • Diagram of AI clients sharing one repository. In your tools, coding agents (Claude Code, Codex, Cursor) connect over hosted MCP or the npx bridge, and AI chats (ChatGPT, Claude, Grok) connect over MCP with OAuth or an API key. Both read from and write to Context Repo, which holds indexed documents, versioned prompts, and organized collections, with Search, Deep Search, and Expand. Both two-way connections are highlighted, so what one client saves, the next one reads.

    Any AI client you already use

    • 29 MCP tools for read, search, and write
    • Deep Search with hierarchical navigation
    • Save from one client, read it from the next

    The AI does the chatting. ChatGPT, Claude, Cursor, Codex, or Claude Code read your grounded context from Context Repo and can save back what they make.

Connect the client you already use. One URL, every assistant.

Claude Desktop

MCP server config. A short JSON block brings all 29 hosted tools into Claude.

Setup guide

Cursor

One-click deep link. Context Repo lands in your Cursor settings with no manual editing.

Setup guide

ChatGPT

ChatGPT app with OAuth. Sign in once and your documents and prompts appear in the conversation.

Connect ChatGPT

Factory

MCP integration for autonomous agents. Droids read and write the same repository you curate.

Other clients

Developer note

Every prompt keeps its version history. When an agent starts behaving differently, I read the diff between versions instead of guessing, and restore the one that worked.

Your questions, answered.

  • Context Repo is an AI context platform that lets you capture, organize, and retrieve prompts, documents, and collections from compatible AI clients. It includes a web dashboard, Chrome Extension, MCP Server with 29 tools, and a ChatGPT App.

Start free

Still copying context between tools?

Put it in one place instead.

Illustrative activity: documents and prompts lighting up as connected AI tools read and write them.