LMGram Connector Documentation

Last updated: June 12, 2026

LMGram is an opt-in human matching platform. Its connector lets AI clients create minimized human match requests, list match requests, and operate lightweight chats after a user connects an LMGram account.

Connection

• ChatGPT MCP endpoint: https://api.lmgram.com/mcp
• Claude MCP endpoint: https://api.lmgram.com/mcp/claude
• Authorization metadata: https://api.lmgram.com/.well-known/oauth-authorization-server
• Protected resource metadata: https://api.lmgram.com/.well-known/oauth-protected-resource
• Privacy policy: https://api.lmgram.com/privacy
• Terms: https://api.lmgram.com/terms

Authentication

The connector uses OAuth 2.0 authorization code flow. LMGram authenticates users with Google sign-in, then issues an LMGram bearer token to the connected AI client. The token only authorizes LMGram connector actions for the connected user.

Supported scopes are profile, match:read, match:write, chat:read, and chat:write.

Tools

• get_current_user: reads the connected LMGram profile.
• request_human_match: creates an opt-in human match request from a minimized summary.
• list_match_requests: reads recent match requests for the connected user.
• start_chat: starts or requests a chat for a match request.
• list_messages: reads messages in a connected user's LMGram chat.
• send_message: sends a message in a connected user's LMGram chat.

Data Minimization

AI clients should send concise summaries and concrete user intent, not raw chat transcripts, secrets, API keys, private documents, or unrelated personal data. LMGram stores the fields needed to operate profiles, match requests, and chats.

Safety Controls

LMGram supports blocking, reporting, chat deletion/hiding, token revocation, and rate limits on public connector surfaces. Users can disconnect the connector from the AI client and can request privacy support through the contact in the privacy policy.