OpenAI

ChatGPT is OpenAI's assistant, and for a lot of teams it's the first place a question gets asked. It answers, drafts, and analyzes, and through connected tools it can act on the systems you point it at.

Your project data is usually the thing you most want to ask about, and the thing a general assistant can't see. Connecting Supabase over the Model Context Protocol closes that gap, giving ChatGPT scoped access to your project so you can ask about a real schema and real rows instead of pasting exports into a chat window.

Once connected, you can:

• Ask what tables exist and how they relate
• Run SQL queries and read the results back
• Read API, Postgres, Auth, and Storage logs
• Check the security and performance advisors
• Pull project URLs and publishable keys
• Search the Supabase docs without leaving the conversation

Getting started:

1. In ChatGPT on the web, open Settings, go to Apps & Connectors, and open Advanced.
2. Turn on developer mode, then choose Create App to add an MCP server.
3. Enter the Supabase MCP server URL: https://mcp.supabase.com/mcp
4. Authorize with Supabase in the browser window that opens, choosing the organization that holds your project.
5. Start a new chat, select the Supabase server from the tools menu, and try asking "What tables are in my database?"

Scope it however you like: pin the connector to a single project, expose only the feature groups you need, or run it read-only, and approve tool calls before they execute.