Stillpeak
Stillpeak · Tools · AI SetupFree

[ AI Setup ]  ·  by dbt Labs

dbt MCP Server

Gives Claude Code or Cursor a live connection to your dbt project — lineage, model health, YAML generation, and 40+ dbt CLI tools inside the agent's context window. No dbt Cloud account required.

Analytics EngineerData Engineer
§ 01 — Install

One command.
That's it.

Runs locally. No account, no backend, no API keys required. Forever free.

TERMINAL

uvx dbt-mcp
View on GitHub→Documentation→

PREREQUISITES

  • ✓uv installed (pip install uv)
  • ✓A dbt project with a valid dbt_project.yml
  • ✓Claude Code, Claude Desktop, or Cursor

STEPS

  1. 01

    Install

    Run uvx dbt-mcp to verify it works. For a permanent install, run pip install dbt-mcp instead.

  2. 02

    Add .mcp.json to your dbt project root

    Create a .mcp.json file pointing to your dbt project directory and dbt executable. This tells Claude Code to connect to the server whenever you open the project.

  3. 03

    Open Claude Code in your dbt project

    Once the config is in place, Claude has access to 40+ tools — full DAG traversal, model health checks, YAML generation, and live dbt CLI commands.

  4. 04

    Try this exact prompt

    "My fct_revenue model failed last night. Trace its full upstream lineage, show me the last test result for each model in the chain, and tell me which one is most likely the source of the issue."

←All tools