What it does
The server connects Supabase projects to Claude and other AI assistants via the Model Context Protocol, exposing database operations, configuration management, and resource control. It supports SQL execution, table schema management, edge function deployment, and storage bucket operations. The server runs in three deployment modes—Supabase cloud (with OAuth), local Supabase CLI, and self-hosted environments—each exposing a different toolset. Access can be scoped to individual projects and restricted to read-only operations via query parameters.
Who it's for
Backend engineers and platform teams who want AI assistants to directly query and modify their Supabase infrastructure. Especially useful for teams doing exploratory database analysis, auto-generating migrations, or having Claude help debug schema and configuration issues without leaving the development environment.
Common use cases
- Run exploratory SQL queries and debug issues against a live Supabase database
- Generate and apply database migrations with AI assistance
- Query and reconfigure storage buckets
- Deploy or test edge functions via LLM
- Troubleshoot schema and access control configuration
Setup pitfalls
- Two secrets detected in the codebase; review Supabase's security best practices before connecting production projects
- Read-only mode is strongly recommended but not default; without it, Claude can execute writes and deletes
- Pre-1.0 status means breaking changes between versions; no CI or test suite in the repo
- Always scope to a specific project using the
project_refparameter to prevent the LLM from accessing all organization projects