Metabase MCP Server: Query Dashboards and Questions from AI

Why Give AI Access to Metabase?

Metabase is the open-source BI tool that teams run on their own infrastructure for SQL queries, dashboards, and data exploration. When analyzing data or debugging metrics, you switch between your editor and Metabase to run saved questions, check dashboard numbers, or explore table schemas. Every tab switch breaks your flow.

An MCP server turns your Metabase instance into structured tools that AI coding assistants (Claude, Cursor, Windsurf) can call directly. Instead of opening Metabase to check why revenue dropped or what yesterday's signups looked like, your AI pulls the numbers in real-time while you work.

What AI Agents Can Do with Metabase Access

With a DataFaucet MCP server pointed at your Metabase instance, AI tools can:

• Run saved questions — execute any saved question and return the results as structured data
• Read dashboards — pull all cards from a dashboard with their current values
• Explore schemas — list databases, tables, columns, and relationships
• Query collections — browse organized folders of questions and dashboards
• Check pulse data — read scheduled report results and alert statuses

Setting Up Metabase as an MCP Server with DataFaucet

1. Open DataFaucet and start a capture session
2. Navigate to your Metabase instance and browse dashboards, run questions, explore the schema browser
3. DataFaucet captures the API calls Metabase makes internally
4. Deploy and connect to your AI coding tool

claude mcp add metabase-data https://your-datafaucet-server.dev/sse

Your AI now has structured tool access to your Metabase data.

Example Queries After Setup

Once connected, ask your AI questions like:

• "What does the revenue dashboard show for this week vs last week?"
• "Run my 'Daily Active Users' saved question and summarize the trend"
• "What tables are in the analytics database and how are they related?"
• "Show me all saved questions in the 'Finance' collection"
• "What did yesterday's scheduled report say about churn?"

Each query returns structured data your AI can use for analysis, reporting, or building features that respond to real metrics.

Why DataFaucet Instead of the Metabase API?

Metabase has a REST API, but it requires generating session tokens, learning endpoint patterns, and handling authentication separately for each tool call. DataFaucet captures from your authenticated session:

No API tokens to generate. No endpoint documentation to read. No session management code.

Common Use Cases

Data-driven development — Ask your AI "what's the current conversion rate?" while building a feature that affects it. No tab switching, no manual lookup.

Incident response — During an outage, ask "what do the error rate dashboards show for the last hour?" and get real numbers without leaving your terminal.

Report automation — Have your AI pull weekly metrics from saved questions and format them for stakeholder updates.

Related: Looker MCP Server for Looker, Grafana MCP Server for Grafana dashboards, PostgreSQL MCP Server for direct database access, Best MCP Servers for the full directory.