Summary
From a question to a saved dashboard in one pass.
Reading Metabase is the easy half. This also writes — create a card from SQL, place it on a dashboard, set the visualization — which is what turns an exploratory query into something a team looks at every morning.
What it is
An MCP server over the Metabase API that reads dashboards, cards and databases, runs queries, and creates new dashboards and cards.
What you get
- Dashboards, saved questions, collections and connected databases all listed
- Tables and their fields listed for any connected database
- Saved questions executed with optional parameters, results returned
- Custom SQL executed against any connected database
- Dashboards created, updated and deleted, and cards positioned on them with optional tab assignment
- New cards created from SQL, and their visualization settings updated
- Collections created, so what you build has somewhere to live
- Resources addressed through metabase:// URIs
Requirements
A Metabase URL plus either an API key or the email and password of a Metabase account. The project recommends the API key for production.
Setup effort
One command plus a key — npx -y @smithery/cli@latest run YOUR_GITHUB_USERNAME/metabase-mcp-server --config {"metabaseUrl":"https://your-metabase-instance.com","metabaseApiKey":"your_api_key","metabasePassword":"","metabaseUserEmail":""}, then supply credentials