Supabase ships an official MCP server at mcp.supabase.com/mcp that lets AI models manage your projects end-to-end — running SQL on the underlying Postgres, inspecting schemas, applying migrations, reading logs and managing Edge Functions. Authenticate with a Supabase personal access token and scope to a single project with ?project_ref=…
https://mcp.supabase.com/mcp
Claude Sonnet 4.5
MCP Playground runs 30+ models on the same workflow: switch anytime, or use Compare mode to run several in parallel and balance quality vs. cost.
Supabase personal access token (app.supabase.com → Account → Access Tokens). Append ?project_ref=<your-ref> to the URL to scope to one project.
How models use it and what it is built for.
The Supabase MCP server wraps the Supabase Management API and the Postgres connection of the project you scope it to. Models can list projects and branches, run read or write SQL against your database, describe schemas and policies, generate and apply migrations, fetch logs from any service (auth, db, storage, functions), and (with the right scopes) deploy or update Edge Functions. It is a one-stop surface for any "build with Supabase" AI workflow.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
List my Supabase projects and tell me which one was last used.
Show me the schema of the public.users table and any RLS policies on it.
Generate and apply a migration that adds a stripe_customer_id column to public.users.
Fetch the last hour of auth logs and summarise any failed login patterns.
This is not a single-model product: you get the same MCP connection with 30+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Supabase
Claude Sonnet 4.5
Claude Sonnet 4.5 is best at multi-step build workflows (schema → migration → query). Haiku 4.5 is faster and cheaper for read-only schema exploration.
Open MCP Agent Studio with the connection pre-filled. Add your token, pick any of 30+ models, and start chatting — no install required.
Try Supabase in Agent StudioCommon questions about connecting, scoping and using it safely.
It is an official MCP server hosted by Supabase at mcp.supabase.com/mcp. It wraps the Supabase Management API and Postgres connection so AI models can manage projects, run SQL, apply migrations and read logs through a single set of tools.
Append ?project_ref=<your-project-ref> to the MCP URL. The server then only operates against that project — useful when an agent should not be able to touch your other Supabase environments.
Yes — it can run write SQL and apply migrations if your access token has those scopes. For exploration we recommend pointing it at a development branch first; Supabase branching makes it cheap to give an agent a throwaway project.
Generic Postgres MCP (e.g. crystaldba/postgres-mcp) only knows about SQL. The Supabase server adds project management, RLS policy inspection, log streaming and Edge Functions — the full Supabase platform, not just the database.
Yes. Any MCP-compatible client works — Claude Desktop, Cursor, Cline, and browser-based tools like MCP Agent Studio. Paste the URL and your personal access token into the client.