orgspecOrg Context Spec, hosted

Your organisation's context, for every AI you use.

Connect your context repository once. Claude, ChatGPT, Cursor and Copilot read it before they work — and propose changes as pull requests a person reviews. The model recommends, a human decides.

Install the GitHub App   Already connected? Sign in to read your context.

How it connects

Claude Code, Cursor, VS Code

Install the App on your context repository and paste one command. The key in it is derived, never stored.

ChatGPT, claude.ai, Gemini, Copilot

Add https://mcp.orgspec.org as a connector. Sign in with GitHub and choose the repository the chat may see.

People

Sign in to read the context in the browser — the overview, every file, and the audit of what it does not say yet.

Nothing is stored on this server: your repository and your keys travel with each request — in plaintext while it is processed — and are gone when it ends. Revoke access any time by uninstalling the App. Rather run it yourself? The same server is MIT licensed: npx orgspec.

Your context repository

Any GitHub repository, private is fine, that follows the Org Context Spec — a handful of Markdown files. No repository yet? Connect an empty one and ask your AI to set it up: it gets the starter templates from the server, interviews you about what you know, and seeds the repository through the same review loop that later maintains it.

Alternative: connect with a fine-grained access token

On GitHub: Settings → Developer settings → Fine-grained tokens → Generate new token.

  1. Repository access: Only select repositories — your context repository alone, so the token can never touch anything else.
  2. Repository permissions: Contents read and write, Pull requests read and write. Both are needed — with Contents alone a proposal is committed to a branch but the pull request fails.

The command is generated in this page; nothing is sent anywhere.

Then run this to connect Claude Code:

Other MCP clients: an HTTP server at with the same headers.