MCP server, by Supabase
Supabase MCP Server
Connect your Supabase projects to Cursor, Claude, Windsurf, and other AI assistants to manage tables, fetch config, and query data.
Quick facts
What it does
The Supabase MCP Server connects Supabase projects to Cursor, Claude, Windsurf, and other AI assistants. The Model Context Protocol standardizes how LLMs talk to external services like Supabase, connecting AI assistants directly with a Supabase project to perform tasks like managing tables, fetching config, and querying data, per the README, which links to the full tool list in Supabase docs.
Most MCP clients use type http with url https://mcp.supabase.com/mcp and the client prompts you to log in to Supabase during setup, choosing the organization that contains the project. Configuration options in the docs include project_ref, read_only, and features URL parameters, and the @supabase/mcp-server-supabase package also exports createToolSchemas() for the Vercel AI SDK MCP client and createSupabaseMcpHandler() for self-hosting the MCP endpoint over HTTP.
Configuration
Configure the MCP client with type http and url https://mcp.supabase.com/mcp, then log in to Supabase when prompted. A custom MCP URL per project can be generated from the MCP connection tab in the Supabase dashboard. Local CLI access is at http://localhost:54321/mcp with a limited tool subset and no OAuth 2.1, per README.
{"mcpServers":{"supabase":{"type":"http","url":"https://mcp.supabase.com/mcp"}}}Source: https://github.com/supabase/mcp. Placeholders only. Replace them in your own environment, and never commit real keys to a repository.
No environment variables are required per the official documentation.
Verified tools and features
Managing tables Config checked
Perform tasks like managing tables in a Supabase project from an AI assistant, per README description of what the MCP connection allows.
Fetching config Config checked
Fetch Supabase project config from an AI assistant, per README description.
Querying data Config checked
Query data in a Supabase project from an AI assistant, per README description.
get_project_url Config checked
Tool name shown in the README AI SDK example, with input { project_id: string } and output { url: string }.
Security notes
- README setup step 1: follow Supabase security best practices to understand the risks of connecting an LLM to Supabase projects and how to mitigate them, before setting up the MCP server.
- README documents read_only and features filtering options for the AI SDK schemas, mirroring URL parameters read_only=true and features, and projectScoped behavior when project_ref is configured. Use read-only and feature restrictions to limit scope where appropriate.
- No env var name is stated in the fetched README setup config, which uses OAuth login via the http endpoint, so none is listed.
An MCP server runs with the permissions of the process you start it from. Scope credentials to the narrowest access the job needs, and review what the client will send before connecting anything sensitive.
Sources and freshness
Last checked 2026-10-07. Verification label: config checked, which means the configuration and feature list trace to the official source above. It does not mean AgentsUse ran the server. Spotted an error? Send a correction. Back to the MCP server index.