Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Data and BI - official dbt Labs server

dbt MCP server

MCP server that gives AI clients access to dbt project metadata, the Semantic Layer, Discovery and Admin APIs, SQL execution and text-to-SQL, plus dbt CLI commands when self-hosted. Available as a self-hosted Python package (uvx dbt-mcp) or a remote server hosted on the dbt platform. It is served from your own instance at https://YOUR_DBT_HOST_URL/api/ai/v1/mcp/ with OAuth or API key auth.

Last verified 15 tools7 sources

Install

Terminal
claude mcp add --transport http dbt https://YOUR_DBT_HOST_URL/api/ai/v1/mcp/
JSON
{
  "mcpServers": {
    "dbt": {
      "url": "https://YOUR_DBT_HOST_URL/api/ai/v1/mcp/"
    }
  }
}

VS Code

JSON
{
  "servers": {
    "dbt": {
      "type": "http",
      "url": "https://YOUR_DBT_HOST_URL/api/ai/v1/mcp/"
    }
  }
}

Generic

JSON
{
  "mcpServers": {
    "dbt": {
      "command": "uvx",
      "args": ["dbt-mcp"],
      "env": {
        "DBT_HOST": "YOUR-ACCESS-URL"
      }
    }
  }
}

Tools

execute_sql
Executes SQL on dbt Platform infrastructure with Semantic Layer support.
text_to_sql
Generates SQL from natural language using project context.
list_metrics
Retrieves all defined metrics.
query_metrics
Executes metric queries with filtering and grouping options.
get_dimensions
Gets dimensions for specified metrics.
get_all_models
Retrieves name and description of all models.
get_lineage
Gets a bounded lineage graph with type, depth and direction filtering.
get_node_details
Gets full details for any dbt resource type.
get_model_health
Gets health signals: run status, test results and upstream source freshness.
build
Executes models, tests, snapshots and seeds in DAG order.
run
Executes models to materialize them in the database.
test
Runs tests to validate data and model integrity.
compile
Generates executable SQL from models, tests and analyses.
trigger_job_run
Triggers a dbt Platform job run, with optional overrides.
get_job_run_error
Gets error and warning details for a job run.

As documented on October 17, 2026. A running server reports its current list through tools/list.

Notes and caveats

  • Remote URL is account-specific: copy it from Account settings > Access URLs > MCP Endpoint URL (for example https://cloud.getdbt.com/api/ai/v1/mcp/).
  • Remote OAuth is in public beta for Starter, Enterprise and Enterprise+ accounts and needs a static subdomain; token auth on remote needs a token with Developer permission (service tokens with that set only on Enterprise plans), and execute_sql requires a PAT.
  • Remote server does not support dbt CLI commands or Codegen; tool availability depends on plan.
  • Snippets shown are the remote OAuth configs; the generic snippet is the self-hosted uvx config (OAuth via dbt platform).
  • Codex is not documented (docs say to consult the client's docs).
  • An experimental .mcpb bundle ships with each release.
  • Latest changelog entry v2.8.0 dated 2026-10-12.

FAQ

What does the dbt MCP server do?

MCP server that gives AI clients access to dbt project metadata, the Semantic Layer, Discovery and Admin APIs, SQL execution and text-to-SQL, plus dbt CLI commands when self-hosted. Available as a self-hosted Python package (uvx dbt-mcp) or a remote server hosted on the dbt platform. The vendor documents 15 tools, including execute_sql, text_to_sql, list_metrics and query_metrics.

How do I add the dbt MCP server to Claude Code?

Run claude mcp add --transport http dbt https://YOUR_DBT_HOST_URL/api/ai/v1/mcp/, copied from dbt Labs's docs. This page also has snippets for Cursor, VS Code and Generic.

Does the dbt MCP server need authentication?

Yes, OAuth or API key. Transport: Streamable HTTP, stdio. Status: GA, last checked October 17, 2026.

Sources

Last verified against 7 dbt Labs pages, then re-checked independently.

Newsletter

New MCP servers, the week they ship

One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.