Supabase MCP
Let the agent read your schema instead of imagining it
Supabase's MCP server exposes project management, schema inspection, SQL execution and log access. The single biggest win is that an agent writing queries can read the actual tables, columns and policies rather than inferring them from your code.
Schema awareness is the point
An agent that cannot see the database guesses at column names and writes queries that fail on the first run. One that can read the schema writes correct SQL immediately, and notices the foreign key you forgot to mention.
Log access matters almost as much. A failing insert usually has a specific policy violation behind it, and reading that error is faster than reasoning about what row level security might be doing.
Use read only mode
The server supports a read only flag and a project scope. Use both. An agent with unrestricted SQL execution against a database holding real data is a risk with no upside during ordinary development work.
For migrations, keep the human in the loop: have the agent write the SQL file and apply it yourself. That is slower by one minute and removes an entire class of accident.
Reach for it when
- the agent writes queries against a schema it should not have to guess
- debugging a row level security policy from the actual error
- keeping types and tables in step without pasting the schema in
Watch out for
- run it read only unless you have a specific reason not to
- a project access token is a real credential, so scope it to one project
- apply migrations yourself rather than letting an agent run them unattended
Compare with these
- MCP serversFree
Chrome DevTools MCP
Let the agent open the page and read what actually rendered
- mcp
- browser
- chrome
Apache-2.0Read - MCP serversFree
Playwright MCP
Browser automation through the accessibility tree instead of pixels
- mcp
- browser
- playwright
Apache-2.0Read - MCP serversFree tier
Context7
Current library documentation, fetched instead of remembered
- mcp
- documentation
- context
MITRead