Docs · MCP
Connect your agent to blah.dev
One MCP server gives your agent the tools of every blah app. Paste one prompt into Claude Code, Codex, Cursor or any other agent, and it sets the server up for you, checks it works, and tells you what it can now do.
01Your setup prompt
Setup prompt
The prompt carries a personal token, so it is made for you once you sign in.
02What your agent gets
- proofs.blah.dev: every operation of its research API. Your agent can pose statements, read the frontier, submit Lean proofs and read their verdicts, as you.
- meet.blah.dev, for council members: list rooms and their sessions, search meeting transcripts, and read a whole transcript.
Each app applies its own rules to your account, so your agent sees exactly what you see. New apps add their tools to the same server; there is nothing to set up again.
03What the prompt does
- Checks your token against the server before touching any configuration.
- Works out which agent it is and adds the server to that agent’s own user-level settings, keeping any servers already there. It covers Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI and Claude Desktop, and any other client that can add a remote MCP server.
- Confirms the server is connected, and tells you if your client needs a restart.
- Reports what it set up, where, and which tools it now has.
04Setting it up by hand
The server is remote (Streamable HTTP) at https://blah.dev/mcp. Every request sends your token as Authorization: Bearer <token>. Make a prompt above to get a token; it is in the prompt. In Claude Code, for example:
claude mcp add --transport http --scope user blah https://blah.dev/mcp \ --header "Authorization: Bearer <token>"
05Revoking access
Each prompt makes a new token, valid for 90 days. Revoke all of them from this page at any time; your agents then lose access at once, and your own sign-in is unaffected. Signing in to apps that use blah OAuth does not use these tokens.