MCP Metabase Server

Community

Comprehensive Metabase access for AI assistants (70+ tools)

How to install ↓ Compare GitHub source repository ↗

Category: Other #83 of 230
Stars: ★ 82 Popular
Transport: stdio (local process)
Runtime: Node.js · Docker
Credentials: API key / credential required
License: MIT
Last commit: 5 months ago
Tools: 95

59FMRS · C

This MCP server provides comprehensive functionality for Metabase via its REST API, including dashboards, cards, databases, tables, collections, and user management. It supports multiple deployment methods and offers tool filtering for controlled access. It is a powerful tool for organizations looking to integrate Metabase with AI workflows, but careful management of API permissions is required to avoid unintended modifications.

Strongest · Documentation 15/20 Weakest · Reliability 6/20

Reliability

6/20

Security and permissions

12/20

Maintenance

12/20

Documentation

15/20

Setup experience

14/20

Why each score

Reliability 6/20: Only the README and repository metadata were supplied; no source code, tests, CI workflows, or server manifest could be inspected, so neither the MCP init/tool-list handshake nor actual tool behavior can be verified. The README's 96-tool claim is internally coherent (27+21+16+17+13+2=96) and a happy path is plausible, but the repository description says 70+ tools, a minor inconsistency. Per static calibration, missing execution evidence lowers the score; 6 is appropriate.
Security and permissions 12/20: No real tokens or secrets appear in the documentation; credentials are passed via METABASE_URL/METABASE_API_KEY environment variables and the server interacts only with the Metabase API, giving a clear data boundary. TOOL_MODE offers all/essential/read/write scoping, but the default is 'all', which exposes permanent destructive operations such as delete_database, replace_table_csv, delete_collection, and revert_dashboard with no confirmation step; the username/password fallback is weaker than an API key. No red lines are triggered (no malware, credential theft, or real secrets), but least privilege is not the default and destructive actions lack confirmation, so 12.
Maintenance 12/20: The repository is not archived, has an MIT license, 3 open issues, an npm version badge, and a clearly identified maintainer (EaseCloud) with a support contact, indicating an active project. However, the supplied material contains no commit history, release cadence, CHANGELOG, or dependency-update evidence to verify governance and versioning. This matches 'active but with governance/versioning gaps' near the 10 anchor, rated 12.
Documentation 15/20: The README is well-layered: quick start, npx/global/Docker installation, authentication and TOOL_MODE matrix, grouped tables describing all 96 tools across six domains, metabase:// resource URIs, development/debugging commands, an issue template, and client config examples for macOS and Windows. Gaps include no troubleshooting section, no per-tool parameter or limit documentation, undisclosed failure modes, and a referenced but unshown .env.example, leaving hidden assumptions. Rated 15 for strong coverage with acknowledged troubleshooting/limitation gaps.
Setup experience 14/20: Setup is clear: a one-line npx command, plus global npm, Docker, and local-build (node dist/index.js) alternatives, with Claude Desktop JSON configuration examples for both macOS and Windows covering npx and local builds. Static calibration caps setup at 15 without verifiable execution evidence, and this review could not confirm the dist/index.js path, docker build, or the existence of .env.example, so a conservative 14 is given.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the networkConnects to a database

Best for

Organizations using Metabase that want AI-driven interactions with their BI platform
Developers building AI agents that need to interact with Metabase data
Teams needing automated Metabase administration and monitoring

Not for

Users not using Metabase
Scenarios requiring real-time, low-latency data queries (this server uses API calls)
Teams looking to avoid maintaining a Metabase instance

Required permissions

Requires an API key or username/password for the Metabase instance
Tools can read and modify dashboards, cards, databases, collections, and users, depending on API token permissions
Schema cache may require local filesystem write access (~/.easecloud/metabase-mcp/)

Risks and side effects

Accidental modification or deletion of Metabase resources if API token permissions are too broad
Write tools can cause irreversible changes if misused
Unencrypted API keys could be exposed
Potential rate limiting or increased load on the Metabase instance due to frequent requests

Setup

Before you start

Runtime：Node.js · Docker

METABASE_URL required Base URL of your Metabase instance, e.g. https://your-metabase-instance.com; required.

METABASE_API_KEY optionalsecret Metabase API key (preferred auth), generated in your Metabase account settings.

METABASE_PASSWORD optionalsecret Metabase password, paired with METABASE_USERNAME as fallback auth.

Other optional settings (2)

METABASE_USERNAME optional Metabase username; fallback auth only when no API key is used.

TOOL_MODE optional Optional; controls tool exposure: all (default), essential, read, write.

Set environment variables METABASE_URL and METABASE_API_KEY (or use username/password).
Run npx @easecloudio/mcp-metabase-server.
Configure your MCP client to point to this command.

claude_desktop_config.json

{
  "mcpServers": {
    "metabase": {
      "command": "npx",
      "args": [
        "@easecloudio/mcp-metabase-server"
      ],
      "env": {
        "METABASE_URL": "https://your-metabase-instance.com",
        "METABASE_API_KEY": "your_metabase_api_key"
      }
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "metabase": {
      "command": "npx",
      "args": [
        "@easecloudio/mcp-metabase-server"
      ],
      "env": {
        "METABASE_URL": "https://your-metabase-instance.com",
        "METABASE_API_KEY": "your_metabase_api_key"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add metabase -e METABASE_URL=https://your-metabase-instance.com -e METABASE_API_KEY=your_metabase_api_key -- npx @easecloudio/mcp-metabase-server

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

The client's tool list should include tools like list_dashboards; running list_dashboards or check_database_health and getting real data back confirms the connection works.

Troubleshooting

Ensure METABASE_URL is accessible from the machine running the server
Verify that the API key is valid and has sufficient permissions
Check that Metabase version is v0.46 or higher, preferably v0.48+
If using Docker, make sure environment variables are passed correctly
Use the MCP Inspector to debug tool calls

Things to try

Once connected, you can ask your AI assistant things like:

List all dashboards in my Metabase instance
Execute card 12 and export the results as CSV
Run a health check on my database connection
Convert this SQL question to an MBQL question and save it as a new card

Tools 95

list_dashboards read-only

List all dashboards

get_dashboard read-only

Get a specific dashboard by ID

create_dashboard writes

Create a new dashboard

update_dashboard writes

Update an existing dashboard

delete_dashboard destructive

Delete / archive a dashboard

copy_dashboard writes

Duplicate a dashboard (shallow or deep copy)

save_dashboard writes

Save a complete dashboard object with nested data

save_dashboard_to_collection writes

Move a dashboard into a specific collection

Show 87 more tools

search_dashboards read-only

Search dashboards by name or description

favorite_dashboard writes

Mark a dashboard as a favourite

unfavorite_dashboard writes

Remove a dashboard from favourites

get_dashboard_cards read-only

Get all cards in a dashboard

add_card_to_dashboard writes

Add a card with positioning

add_text_block writes

Add a text or heading block to a dashboard

remove_card_from_dashboard writes

Remove a card from a dashboard

update_dashboard_card writes

Update card position, size, and settings

update_dashboard_cards writes

Bulk-replace all cards on a dashboard

update_dashcard writes

Update a specific dashcard's properties

execute_dashboard_card read-only

Execute a specific dashcard and return results

create_dashboard_public_link writes

Create a public sharing link

delete_dashboard_public_link writes

Remove the public link

list_public_dashboards read-only

List dashboards with public links

list_embeddable_dashboards read-only

List dashboards enabled for embedding

get_dashboard_revisions read-only

Get revision history (audit trail)

revert_dashboard writes

Revert to a previous revision

get_dashboard_related read-only

Get related content suggestions

get_dashboard_queries read-only

Extract all card queries with resolved field IDs

list_cards read-only

List all questions / cards

get_card read-only

Get a card by ID (includes full SQL / MBQL query)

create_card writes

Create a new question

update_card writes

Update an existing question

delete_card destructive

Delete / archive a question

copy_card writes

Duplicate a card

move_cards writes

Move one or more cards to a different collection

move_cards_to_collection writes

Bulk-move cards from a source collection

execute_card read-only

Run a card and return results

execute_pivot_card_query read-only

Run a card formatted as a pivot table

export_card_result read-only

Export results as CSV or JSON

get_card_param_values read-only

Get available values for a card parameter

search_card_param_values read-only

Search / filter parameter values

get_card_param_remapping read-only

Get how parameter values are remapped for display

get_card_query_metadata read-only

Get column types and display names

get_card_dashboards read-only

List dashboards containing this card

get_card_series read-only

Get series data or related card suggestions

create_card_public_link writes

Create a public sharing link

delete_card_public_link writes

Remove the public link

list_public_cards read-only

List cards with public links

list_embeddable_cards read-only

List cards enabled for embedding

list_databases read-only

List all database connections

get_database read-only

Get a specific database by ID

create_database_connection writes

Create a new database connection

update_database writes

Update a database connection's config or credentials

delete_database destructive

Permanently remove a database connection

add_sample_database writes

Add the built-in H2 sample database with demo data

execute_query read-only

Execute a SQL query against a database

get_database_schema read-only

Get schema information

get_database_tables read-only

Get all tables in a database

get_database_metadata read-only

Get full metadata including all tables and field IDs

list_database_schemas read-only

List all schemas within a database

sync_database_schema writes

Trigger a schema metadata sync

get_database_sync_status read-only

Check sync status

check_database_health read-only

Check connection health

validate_database read-only

Validate connection settings before saving

test_database_connection read-only

Test an existing connection

list_tables read-only

List tables (optionally filtered by database)

get_table read-only

Get table metadata by ID

get_table_metadata read-only

Get full query metadata including field IDs and types

get_table_fks read-only

Get foreign key relationships

get_table_related read-only

Find related tables via FK relationships

get_table_data read-only

Get a sample data preview from a table

get_field_id read-only

Look up a field ID by table ID + column name (returns MBQL ref)

get_card_table_fks read-only

Get FK relationships for a card's virtual table

get_card_table_query_metadata read-only

Get query metadata for a card's virtual table

update_table writes

Update display name, description, or visibility

update_tables writes

Bulk-update multiple tables with the same config

reorder_table_fields writes

Change the display order of fields

sync_table_schema writes

Trigger a schema sync for a specific table

rescan_table_field_values writes

Rescan field values (updates filter dropdowns)

discard_table_field_values writes

Discard cached field values

append_csv_to_table writes

Append new rows from CSV content

replace_table_csv destructive

Replace all table data with new CSV content

list_collections read-only

List all collections

create_collection writes

Create a new collection

get_collection read-only

Get collection details

update_collection writes

Update name, description, color, or parent

delete_collection destructive

Delete a collection and its contents

get_collection_items read-only

List cards, dashboards, and sub-collections

move_to_collection writes

Move a card or dashboard to a different collection

list_users read-only

List all users

create_user writes

Create a new user

list_permission_groups read-only

List all permission groups

create_permission_group writes

Create a new permission group

search_content read-only

Search across all Metabase content

get_schema_cache read-only

Return cached schema (auto-fetches if missing or stale)

refresh_schema_cache writes

Force-refresh cache for one or all databases

Use cases

Automating dashboard, card, and database management in Metabase

Querying and manipulating Metabase data via natural language

Integrating AI assistants into BI workflows

Supported clients

Claude Desktop

Listed from the project's documentation, not tested by this site.

Overview

A Model Context Protocol (MCP) server for Metabase that gives AI assistants full access to your analytics platform — dashboards, cards, databases, tables, collections, and more. Developed and maintained by EaseCloud.

Similar servers

Hevy MCP 77 · B

Manage your Hevy workout data from AI assistants via MCP.

★ 517 · Tools 26 Compare with this →

Firecrawl MCP Server 75 · B

Firecrawl's official MCP server for web search, scraping, and structured extraction for AI agents

★ 7.6k · Tools 11 Compare with this →

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →

Compartment 75 · B

Encrypted, fully offline long-term memory for AI agents

★ 583 · Tools 14 Compare with this →

Compare with other MCP servers

Source revision 464fc93edd22 Data synced 2026-10-19 Read the FMRS scoring method