supabase/mcp

What it solves

This project provides a standardized way for AI assistants (like Cursor, Claude, and Windsurf) to interact directly with Supabase projects. It eliminates the need for manual data entry or custom integrations by allowing LLMs to manage tables, fetch configurations, and query data within a Supabase environment.

How it works

It implements the Model Context Protocol (MCP), which acts as a bridge between Large Language Models and external services. The server can be hosted by Supabase or self-hosted via the @supabase/mcp-server-supabase package. It exposes a set of tools that AI clients can call to perform actions in Supabase, with support for project-scoped access, read-only modes, and specific feature group restrictions.

Who it’s for

• AI-assisted developers who want their AI coding assistants to have direct access to their database and project settings.
• Developers using the Vercel AI SDK who need typed tool schemas for Supabase interactions.
• Developers who want to self-host their own MCP endpoint for Supabase tools.

Highlights

• Broad Compatibility: Works with major AI assistants like Claude, Cursor, and Windsurf.
• Flexible Deployment: Supports cloud-hosted, local CLI, and self-hosted environments.
• Vercel AI SDK Integration: Provides createToolSchemas() for client-side validation and TypeScript types.
• Granular Control: Offers configuration options to restrict tools to read-only access or specific feature groups (e.g., database or docs).