ServicesAboutNotesContact Get in touch →
EN FR
Note

dbt MCP Server Setup

A reading path through connecting dbt to AI assistants via MCP — choosing between local and remote modes, tool capabilities, configuration, and safety.

Planted
mcpdbtclaude codeaidata engineering

The dbt MCP server connects your dbt project to AI assistants like Claude Desktop and Claude Code. You ask about models, lineage, and metrics through conversation. With the local server, you can also run dbt commands — build, test, compile — without leaving your chat window.

This hub collects four notes covering the decision framework, tool capabilities, setup procedures, and safety considerations. They assume familiarity with MCP fundamentals and basic dbt usage.

Prerequisites

  • A dbt installation (Core, Fusion Engine, or Cloud CLI)
  • A dbt project with dbt_project.yml
  • The uv package manager from Astral
  • An MCP client: Claude Desktop, Claude Code, or similar
  • (Optional) A dbt Cloud account for Semantic Layer and job management

Reading Order

Start with dbt MCP Server: Local vs Remote to decide which deployment mode fits your needs. The quick version: data engineers want the local server for CLI access; analysts who only need Semantic Layer queries can use the remote server.

dbt MCP Server Tool Reference catalogs all 20+ tools across four categories — CLI commands, metadata discovery, Semantic Layer queries, and job management. Use this to understand what’s possible and match tools to your workflow.

dbt MCP Server Setup and Configuration walks through installation, environment variables, feature toggles, and client configuration for both Claude Code and Claude Desktop.

dbt MCP Server Safety Considerations covers the risks of giving an AI assistant dbt CLI access and practical mitigations — development profiles, feature toggles, safety hooks, and credential scoping.

  • BigQuery MCP Server Setup — The equivalent hub for connecting BigQuery directly to AI assistants. The dbt MCP server and BigQuery MCP server complement each other: dbt MCP for model-aware operations, BigQuery MCP for direct SQL execution and schema exploration.
  • Advanced Claude Code Workflows for dbt — Broader Claude Code configuration for dbt projects, including hooks, slash commands, TDD, and debugging workflows.
  • MCP Protocol Architecture — How the protocol works at the architecture level, for readers who want to understand the foundations.