Supabase MCP — Database & Auth for AI Agents

What it is

Supabase MCP is the official Model Context Protocol server built by Supabase. It lets AI coding assistants like Claude Code or Cursor interact with your Supabase project directly -- running SQL queries, managing authentication, accessing file storage, calling edge functions, and subscribing to real-time changes.

It targets developers who build applications on Supabase and want their AI agent to understand their database schema, write queries, debug issues, and manage backend resources without context-switching to the Supabase dashboard.

How it saves time or tokens

Instead of manually describing your database schema to the AI or copy-pasting query results, the MCP server gives the agent direct access. The agent can inspect your tables, understand relationships, write and test SQL queries, and even create database migrations. This eliminates the schema-description overhead that wastes both your time and prompt tokens.

How to use

1. Add the Supabase MCP server to your editor configuration:

{
  "mcpServers": {
    "supabase": {
      "command": "npx",
      "args": [
        "-y",
        "@supabase/mcp-server-supabase@latest",
        "--access-token",
        "sbp_YOUR_TOKEN"
      ]
    }
  }
}

1. Generate a personal access token at https://supabase.com/dashboard/account/tokens.
2. Restart your AI editor.
3. Ask the agent to interact with your Supabase project:

'Show me the schema of the users table and suggest an index for the email column.'

Example

Related on TokRepo

• Supabase MCP integration -- deep-dive into Supabase MCP setup
• AI tools for database -- database tools for AI agents

Common pitfalls

• The personal access token grants broad access to your Supabase project. Use a token scoped to development projects, not production, when experimenting with AI agents.
• Row Level Security (RLS) policies apply to SQL queries run through the MCP server. If your queries return empty results, check that the token's role has the necessary permissions.
• Schema changes made by the agent (CREATE TABLE, ALTER TABLE) are applied immediately. Review generated migrations before letting the agent execute them.