Dbt MCP is a server tool by Dbt Labs, built with Python 3.12+, designed to integrate AI applications with dbt-managed data assets. It functions as a Model Context Protocol server, bridging AI assistants with dbt projects and the dbt Semantic Layer. This server connects to dbt Core, dbt Fusion, and dbt Platform. Dbt MCP is distributed as an experimental Model Context Protocol Bundle, via uvx for self-hosted installations, as a Python package on PyPI, and from its GitHub repository. It supports various AI clients, facilitating conversational data access and AI-assisted development workflows.

What does Dbt MCP connect to?

Dbt MCP serves as a Model Context Protocol server, connecting to dbt Core, dbt Fusion, and dbt Platform. Its primary role is to integrate AI applications with dbt-managed data assets. It acts as a bridge between AI assistants and either dbt projects directly or the dbt Semantic Layer. This standardized framework allows AI applications to access and interact with dbt environments for various data-related tasks and insights.

Core features and tools

Dbt MCP exposes a range of callable tools, which vary between self-hosted and remote server types. These include SQL Tools like execute_sql and text_to_sql, and Semantic Layer Tools such as query_metrics and list_metrics. Discovery Tools offer functions like get_all_models and get_lineage_dev. Self-hosted servers gain dbt CLI Tools including build, run, and compile. Administrative API Tools provide list_jobs_runs and retry_job_run. Additionally, Fusion Tools expose dbt Fusion's column-level lineage, complemented by MCP Server Metadata Tools for version information.

Who is it for?

Dbt MCP serves a broad audience, including AI agents, AI applications, developers, and data teams. Analytics engineers, API developers, and backend engineers can utilize its capabilities. Non-technical stakeholders, business users, and analysts also benefit from its data access features. It enables AI assistants to interact with dbt projects and the dbt Semantic Layer. Skip this if your AI applications do not require direct, programmatic interaction with dbt-managed data assets.

How to get started

Access is granted through a standard OAuth authorization flow or token-based authentication, which can utilize a Personal Access Token or service token. Configuration for the self-hosted server is managed via environment variables, and the remote server specifically uses an HTTP connection for transport, requiring client configuration with a valid MCP URL and appropriate headers.

Use cases

• Enabling AI agents and business users to query dbt models, metrics, and metadata using natural language for consistent and governed results.
• Allowing AI agents to execute dbt CLI commands, manage job runs, and debug failed runs for agentic automation of dbt workflows.
• Providing AI assistants with project context, generating SQL from natural language, compiling models, validating project file syntax, and generating documentation to aid developers.

Dbt MCP provides a critical bridge for integrating AI applications with dbt data assets, offering a diverse set of tools for data interaction and workflow automation. Its support for various authentication methods and transport protocols ensures adaptability for both remote and self-hosted environments. The server empowers AI agents and developers to interact more deeply with dbt projects.