Supabase MCP Server — Database, Auth & Storage for AI Agents
Install the Supabase MCP server to give your AI agent direct access to your Supabase project — query Postgres tables, manage auth users, upload to Storage, and call Edge Functions, all from Claude Code or Cursor.
Install Supabase MCP Server
{
"mcpServers": {
"supabase": {
"command": "npx",
"args": ["-y", "@supabase/mcp-server-supabase", "--access-token", "sbp_your_token"]
}
}
}Supabase MCP — Database & Auth for AI Agents
Official Supabase MCP server. Connect AI agents to Supabase for database queries, auth management, storage, edge functions, and real-time subscriptions. 2.6K+ stars.
Supabase MCP — Postgres + Auth Server for AI Agents
MCP server for Supabase that gives AI agents access to PostgreSQL databases, authentication, storage, and edge functions. Query data, manage users, and build full-stack features. 4,000+ stars.
MCP.so — Open-Source MCP Server Directory
Self-host your own MCP server directory with Next.js and Supabase. Browse, search, and submit community MCP servers. Powers mcp.so. Apache-2.0, 2,000+ stars.
Your full-stack backend, one MCP away
The Supabase MCP server goes beyond a simple database connector. Unlike the generic PostgreSQL MCP server which gives you raw SQL access, the Supabase server exposes the full Supabase platform — Postgres tables, Row Level Security policies, Auth users, Storage buckets, Edge Functions, and realtime subscriptions — as individual MCP tools the AI can reason about.
This matters because Supabase is a complete backend. When a developer asks "add a file upload feature," the AI agent needs to create a storage bucket, set RLS policies, write the upload API endpoint, and update the frontend. With the Supabase MCP server, the agent can do all of this in one conversation because it has visibility into every layer — no more "I've written the code but you'll need to create the bucket manually."
The server authenticates via your personal access token (not project API keys), which means it can manage multiple Supabase projects from a single connection. For teams building full-stack applications, pair it with the GitHub MCP server for automated PRs that include both frontend code and backend migrations. Browse the full list of database AI tools on TokRepo for alternatives like Neon, PlanetScale, and raw Postgres connectors.
Supabase gives your AI agent the same backend superpowers it gives developers — database, auth, storage, and functions in one platform.
Frequently Asked Questions
What is the Supabase MCP server?+
The official Supabase MCP server (@supabase/mcp-server-supabase) lets AI agents interact with your Supabase projects. It exposes tools for querying Postgres, managing tables and RLS policies, working with Auth users, Storage buckets, and Edge Functions. It authenticates with your personal access token and can manage multiple projects.
How do I get a Supabase access token?+
Go to supabase.com/dashboard/account/tokens, click 'Generate new token', give it a name, and copy the token (starts with sbp_). This is a personal access token that grants the same permissions as your Supabase dashboard account — keep it secure and never commit it to git.
How does this differ from the PostgreSQL MCP server?+
The PostgreSQL MCP server gives you raw SQL access to any Postgres database. The Supabase MCP server wraps the full Supabase Management API — including project creation, RLS policies, Auth configuration, Storage management, and Edge Functions. Use the Postgres MCP server when you only need SQL queries; use the Supabase server when you want the agent to manage your entire backend platform.
Can the Supabase MCP server create new tables and migrations?+
Yes. The server can create tables, add columns, set up foreign keys, and configure Row Level Security policies. For production use, we recommend reviewing AI-generated schema changes before applying them — the agent should propose migrations in a development project first, then you promote to production after review.
Is the Supabase MCP server safe for production projects?+
The server uses your personal access token, which has full dashboard-level permissions. For production safety: (1) Use a separate Supabase account with limited project access if possible. (2) Always review AI-proposed schema changes before applying. (3) Point the agent at a staging project for development work. (4) The server doesn't bypass RLS — data queries still respect your security policies.