Metabase ships its own MCP endpoint — search your BI content, build and run queries, and save questions and dashboards without leaving the chat.
Metabase MCP server is a hosted integration for AI assistants that speak the Model Context Protocol. Metabase's built-in MCP server exposes search, query construction, SQL execution and question or dashboard authoring over Streamable HTTP, scoped to the user's own permissions.
Metabase's built-in MCP server exposes search, query construction, SQL execution and question or dashboard authoring over Streamable HTTP, scoped to the user's own permissions.
Being a remote server, there is no local install. You register the endpoint with your client, authorise it once, and the tools appear.
Once Metabase is connected, these are the calls the assistant has available:
search — Find tables, metrics, cards, dashboards and collections by keyword or natural-language queryread_resource — Read up to five Metabase entities at once by metabase:// URI, covering database, schema, table, collection, question, dashboard, metric and transformconstruct_query — Build a query against a table or metric and return an opaque query handle for later execution or savingconstruct_native_query — Build a native SQL query for a database and return a handle intended for saving as a card; it does not execute the SQLquery — Query a table or metric directly, with continuation-token paginationexecute_query — Run a previously constructed query handle and return rows plus column metadataexecute_sql — Run raw SQL against a database, subject to native-query permission and the instance-level enable settingexecute_question — Run a saved question by id under the caller's permissions and return its rows and columnscreate_question — Save a constructed query as a named question (card)update_question — Patch a saved question — rename it, move it between collections, replace its query, or archive itcreate_metric — Save a query with one aggregation as a reusable metricupdate_metric — Patch, move or archive an existing metricA running Metabase instance whose version includes the built-in MCP handler, reachable over HTTPS, plus a Metabase login. Raw SQL execution additionally needs native-query permission and the mcp-execute-sql-enabled instance setting turned on.
This sits in the database access group, where several servers overlap in what they claim to do but differ sharply once you actually set them up. Metabase's toolset — search, read_resource, construct_query and 12 more — is a fair guide to whether it matches your workflow. It is maintained by Metabase rather than a third party, which usually means auth handling and breaking changes are dealt with faster.
This entry was verified against Metabase's own documentation before publication; SyncDev keeps the directory reviewed rather than auto-generated.
Ask a question in plain language, let the model find the right table or curated metric, construct the query and return rows — all inside your existing permission model.
Once an exploration lands somewhere useful, save it as a question and drop it onto a new dashboard so the rest of the team inherits the work.
Walk databases, schemas and tables by URI to answer 'where does this number actually come from?' before anyone writes a query.
Issue a token limited to search, resource read and query execution scopes so an agent can answer questions but cannot create, edit or archive anything.
| Tool | What it does |
|---|---|
| search | Find tables, metrics, cards, dashboards and collections by keyword or natural-language query. |
| read_resource | Read up to five Metabase entities at once by metabase:// URI, covering database, schema, table, collection, question, dashboard, metric and transform navigation. |
| construct_query | Build a query against a table or metric and return an opaque query handle for later execution or saving. |
| construct_native_query | Build a native SQL query for a database and return a handle intended for saving as a card; it does not execute the SQL. |
| query | Query a table or metric directly, with continuation-token pagination. |
| execute_query | Run a previously constructed query handle and return rows plus column metadata. |
| execute_sql | Run raw SQL against a database, subject to native-query permission and the instance-level enable setting. |
| execute_question | Run a saved question by id under the caller's permissions and return its rows and columns. |
| create_question | Save a constructed query as a named question (card). |
| update_question | Patch a saved question — rename it, move it between collections, replace its query, or archive it. |
| create_metric | Save a query with one aggregation as a reusable metric. |
| update_metric | Patch, move or archive an existing metric. |
| create_dashboard | Create a dashboard, optionally pre-populated with saved questions auto-positioned on the grid. |
| update_dashboard | Update dashboard name, description or collection, or archive it. |
| create_collection | Create a collection, optionally nested under a parent collection. |
claude mcp add metabase https://your-metabase.example.com/api/metabase-mcp --transport streamable-httpA running Metabase instance whose version includes the built-in MCP handler, reachable over HTTPS, plus a Metabase login. Raw SQL execution additionally needs native-query permission and the mcp-execute-sql-enabled instance setting turned on.
Read-only SQL access to Postgres — let your assistant inspect schemas and answer questions from real data.
Manage your whole Supabase project in conversation — database, auth, storage, Edge Functions and branches.
Query, modify and analyse local SQLite databases in conversation — the fastest way to chat with a data file.
Official MongoDB server covering data, schemas and Atlas management — from find queries to spinning up clusters.
Serverless Postgres with database branching — point your assistant at Neon and let it work on disposable copies.
Official Redis server — search keys, inspect data structures and manage caches in natural language.