mcp-metabase-server

MCP server that connects AI agents to Metabase — query dashboards, run questions, and surface data through Claude.

MCP

Metabase

Claude

TypeScript

Open Source

What it does

mcp-metabase-server implements the Model Context Protocol (MCP) to expose your Metabase instance as a set of tools and resources that AI agents — including Claude — can interact with directly.

Once configured, an AI agent can list dashboards, execute saved questions, run ad-hoc SQL queries against connected databases, and retrieve results as structured data — without requiring a human to copy-paste query outputs into a chat window.

Built for data teams who want to give AI assistants real-time access to their analytics layer without building custom API integrations from scratch.

Quick start

Install

npm install -g mcp-metabase-server

Configure your Metabase connection

export METABASE_URL="https://your-metabase.example.com"
export METABASE_USERNAME="admin@example.com"
export METABASE_PASSWORD="your-password"

Start the MCP server

mcp-metabase-server start

Key features

List and browse Metabase dashboards and saved questions via MCP resource endpoints
Execute saved questions and return results as structured JSON for downstream AI processing
Run ad-hoc SQL queries against any Metabase-connected database with result streaming
Compatible with Claude Desktop, Claude API tool use, and any MCP-compliant AI client

Building AI on your data stack?

MCP servers are one piece of an AI data pipeline. Our AI & ML consulting team architects the full stack: data ingestion, model selection, RAG pipelines, and production deployment.

Contributing

Contributions welcome — particularly around additional Metabase API coverage and authentication methods. Open an issue to discuss before submitting large PRs.

Take it further

Building AI on your data stack? We architect the full pipeline.

This OSS tool

Exposes Metabase as MCP tools for AI agents
Handles authentication and query execution
Returns structured results to the AI client
Self-hosted, no vendor lock-in

Our managed service

End-to-end AI data pipeline design and implementation
Model selection and fine-tuning for your specific data domain
RAG architecture for internal knowledge bases and analytics
Production deployment with monitoring and cost guardrails

EaseCloud Terminal

→|

What it does

mcp-metabase-server implements the Model Context Protocol (MCP) to expose your Metabase instance as a set of tools and resources that AI agents — including Claude — can interact with directly.

Built for data teams who want to give AI assistants real-time access to their analytics layer without building custom API integrations from scratch.

Key features

List and browse Metabase dashboards and saved questions via MCP resource endpoints

Execute saved questions and return results as structured JSON for downstream AI processing

Run ad-hoc SQL queries against any Metabase-connected database with result streaming

Compatible with Claude Desktop, Claude API tool use, and any MCP-compliant AI client