dbt MCP Server

Name: dbt MCP Server
Author: dbt Labs

dbt Labs' server for running dbt commands, querying metrics and reading project metadata

Official · dbt Labs Python Local · stdio Data and analytics Databases Developer tools

GitHub PyPI Website

81 popularity score ★ 617 ↓ 76k/mo

Connect your client

Local (PyPI)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs DBT_HOST. Replace each <NAME> with your value.

Claude CodeTerminal

claude mcp add dbt -e DBT_HOST=<DBT_HOST> -- uvx dbt-mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "dbt": {
      "command": "uvx",
      "args": [
        "dbt-mcp"
      ],
      "env": {
        "DBT_HOST": "<DBT_HOST>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "dbt": {
      "command": "uvx",
      "args": [
        "dbt-mcp"
      ],
      "env": {
        "DBT_HOST": "<DBT_HOST>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "dbt": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "dbt-mcp"
      ],
      "env": {
        "DBT_HOST": "<DBT_HOST>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.dbt]
command = "uvx"
args = ["dbt-mcp"]

[mcp_servers.dbt.env]
DBT_HOST = "<DBT_HOST>"

Gemini CLITerminal

gemini mcp add --scope user -e DBT_HOST=<DBT_HOST> dbt uvx -- dbt-mcp

Settings 12

`DBT_HOST`	Required: dbt platform access URL hostname (e.g. abc123.us1.dbt.com); turns on OAuth sign-in and the platform tools. For CLI-only use, set DBT_PROJECT_DIR and DBT_PATH instead.
`DBT_PROJECT_DIR`	Optional: Absolute path to the folder holding dbt_project.yml; needed for the CLI tools.
`DBT_PATH`	Optional: Absolute path to the dbt executable; defaults to dbt on the PATH.
`DBT_TOKEN`	Optional: Personal access or service token, instead of OAuth; execute_sql needs a personal token.
`DBT_PROD_ENV_ID`	Optional: Production environment ID, for Discovery and Semantic Layer tools with a token.
`DBT_ACCOUNT_ID`	Optional: Account ID, for Admin API tools and personal-token sign-in.
`DBT_DEV_ENV_ID`	Optional: Development environment ID, for execute_sql.
`DBT_USER_ID`	Optional: dbt platform user ID, for execute_sql.
`DBT_CLI_TIMEOUT`	Optional: Seconds before a dbt command times out; default 60.
`DISABLE_DBT_CODEGEN`	Optional: Set to false to enable the Codegen tools, off by default.
`DISABLE_TOOLS`	Optional: Comma-separated tool names to hide.
`DISABLE_DBT_CLI`	Optional: Set to true to turn off the dbt CLI tools.

What it does

dbt Labs publishes this server so an MCP client can work with a dbt project. Run locally with uvx, it can call the dbt executable on your machine (build, run, test, compile, show and others) against a project folder, and, when connected to a dbt platform account, read project metadata through the Discovery API, query metrics through the Semantic Layer, run or generate SQL, and list, start, retry or cancel jobs through the Administrative API. It also searches and fetches pages from docs.getdbt.com.

Toolsets switch themselves off when their settings are missing rather than failing: with only DBT_PROJECT_DIR and DBT_PATH it offers the CLI tools; with DBT_HOST it signs in to dbt platform by OAuth and adds the platform tools. Each toolset can also be turned off with a DISABLE_* setting, or single tools with DISABLE_TOOLS.

The settings file at v2.7.0 shows defaults the README does not list: the Codegen tools (DISABLE_DBT_CODEGEN) and the server-metadata tools (DISABLE_MCP_SERVER_METADATA) are off unless enabled, DBT_PATH falls back to plain "dbt" on the PATH, CLI commands time out after 60 seconds (DBT_CLI_TIMEOUT), and Semantic Layer responses are cut at 16,000 characters (DBT_MCP_SL_MAX_RESPONSE_CHARS). By default it also loads interactive result views from a dbt-owned CDN, which DISABLE_MCP_APPS turns off.

dbt also runs a hosted version on dbt platform for reading metrics, metadata and lineage without installing anything; it lacks the CLI and Codegen tools.

What it needs

Runtime	Python

uv (the launch command is uvx). For the CLI tools: a local dbt project and a dbt executable (dbt v1, dbt v2/Fusion or the dbt platform CLI). For Semantic Layer, Discovery, Admin API and SQL tools: a dbt platform account; OAuth needs a Starter, Enterprise or Enterprise+ account with a static subdomain, otherwise a token. Local column lineage needs the dbt VS Code extension.

Limits

Platform tools need a dbt platform account, and which ones work depends on the plan's API access. execute_sql works only with a personal access token, not a service token. text_to_sql uses dbt Copilot actions and stops when the allotment runs out. CLI commands time out after 60 seconds by default. The hosted server has no CLI or Codegen tools. search is alpha and not generally available; eleven Discovery tools are deprecated in favour of get_node_details and get_lineage. The MCP bundle (dbt-mcp.mcpb) is marked experimental.

Who it suits

Suits analytics engineers who already use dbt and want an assistant to build, test and inspect their project, and teams on dbt platform who want metrics, lineage and job status in chat. It runs dbt commands that change warehouse objects, so it suits people who trust their client with that; for read-only metric and metadata questions without a local install, dbt's hosted server is the narrower choice. Not for people without a dbt project: it works on nothing else.

Tools 60 of 62

execute_sql: Runs SQL on dbt platform infrastructure, with Semantic Layer support
text_to_sql: Turns a plain-language question into SQL using project context
get_dimension_values: Lists distinct values of a dimension, optionally for given metrics
get_dimensions: Lists the dimensions available for chosen metrics
get_entities: Lists the entities available for chosen metrics
get_metrics_compiled_sql: Shows the SQL a metric query compiles to, without running it
list_metrics: Lists every metric defined in the Semantic Layer
list_saved_queries: Lists the saved queries in the Semantic Layer
query_metrics: Runs a metric query with filters, grouping and ordering
get_all_macros: Lists macros, optionally per package or just package names
get_all_models: Lists every model with its description
get_all_sources: Lists sources and their freshness, optionally by source name
get_exposures: Lists exposures such as downstream dashboards and apps
get_lineage: Returns a lineage graph limited by type, depth and direction
get_mart_models: Lists the mart models
get_model_health: Reports a model's last run status, test results and source freshness
get_model_performance: Shows a model's run history, optionally with test results
get_node_details: Returns full details of any resource: model, source, test, seed, macro and more
get_related_models: Finds models similar to a given one by semantic search
search: Alpha search across project resources; not generally available
get_model_details: Deprecated; replaced by get_node_details
get_model_parents: Deprecated; replaced by get_lineage
get_model_children: Deprecated; replaced by get_lineage
get_source_details: Deprecated; replaced by get_node_details
get_exposure_details: Deprecated; replaced by get_node_details
get_macro_details: Deprecated; replaced by get_node_details
get_seed_details: Deprecated; replaced by get_node_details
get_semantic_model_details: Deprecated; replaced by get_node_details
get_snapshot_details: Deprecated; replaced by get_node_details
get_test_details: Deprecated; replaced by get_node_details
build: Runs dbt build: models, tests, snapshots and seeds in dependency order
clone: Runs dbt clone to copy selected nodes from a saved state into the target schema
compile: Runs dbt compile to render models and tests into SQL
docs: Runs dbt docs to generate project documentation
get_lineage_dev: Reads lineage from the local manifest.json
get_node_details_dev: Reads node details from the local manifest.json
list: Runs dbt list to list project resources by type or selector
parse: Runs dbt parse to check project files
run: Runs dbt run to build models in the warehouse
show: Runs dbt show to execute SQL and return rows
test: Runs dbt test against models and data
cancel_job_run: Cancels a job run in progress
get_job_details: Shows a job's triggers, schedule and commands
get_job_run_artifacts: Downloads one artifact from a job run, with optional jq filter
get_job_run_details: Shows a run's status, timing, steps and artifacts
get_job_run_error: Shows the errors or warnings of a job run
list_job_run_artifacts: Lists the artifacts a job run produced
list_jobs: Lists jobs, optionally by project or environment
list_jobs_runs: Lists job runs, filtered by job or status
list_projects: Lists the projects in the dbt platform account
retry_job_run: Retries a failed job run
trigger_job_run: Starts a job run, optionally overriding branch or schema
generate_model_yaml: Writes model YAML with columns, optionally inheriting descriptions
generate_source: Writes source YAML by inspecting warehouse schemas
generate_staging_model: Writes staging model SQL for a source table
fusion.compile_sql: Compiles SQL in project context through dbt platform
fusion.get_column_lineage: Traces column-level lineage through dbt platform
get_column_lineage: Traces column-level lineage locally through the dbt VS Code extension's language server
search_product_docs: Searches docs.getdbt.com and returns matching page titles and links
get_product_doc_pages: Fetches whole docs.getdbt.com pages as Markdown

Popularity

81 popularity score, on our list (the bar is 36)

★ 617 ↓ 76k/mo

GitHub stars	617	33.5 points
Stars in 30 days	+10	5.2 points
PyPI downloads, last month	75,757	34.2 points
Official server	Yes	8 points

Measured 2026-11-07. How the score works.

Where it is published

Code	dbt-labs/dbt-mcp ★ 617
PyPI	dbt-mcp 2.10.1, 2026-11-03

What it can reach

Runs	On your machine
Reaches	Your code, Files on your machine, A database, Cloud resources, The internet
Changes things	Yes: it can write, send, create or delete
Sign-in	OAuth sign-in
Licence	Apache-2.0 (GitHub)
Last code change	2026-11-05

Runs on your machine and calls the dbt executable named in DBT_PATH against the project in DBT_PROJECT_DIR; build, run, clone and similar tools create or replace warehouse tables, and show or execute_sql run SQL. dbt's README warns these commands can change models, sources and warehouse objects. With DBT_HOST set it signs in to dbt platform by OAuth (or uses DBT_TOKEN) and can start, retry and cancel jobs. Tokens set inline sit in the client config file. It honours DO_NOT_TRACK for usage tracking, and by default fetches UI views from cloud-ui.cdn.getdbt.com unless DISABLE_MCP_APPS is set.

Over time

GitHub stars 617 +10 (+2 %)

617607

9 Oct 20267 Nov 2026

PyPI downloads a month 76k +3.3k (+5 %)

108k72k

12 May 20267 Nov 2026

Before the site measured it, read from pypistats' daily download counts.

Popularity score 81 +13 (+20 %)

8168

9 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-10-14.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![dbt MCP Server on EveryMCPServer](https://everymcpserver.com/badge/dbt-mcp.svg)](https://everymcpserver.com/servers/dbt-mcp)

Connect your client

Add in one click

Or add it by command or config

Settings 12

What it does

What it needs

Limits

Who it suits

Tools 60 of 62

Similar servers

DBX MCP Server

Convex MCP Server

Dremio MCP Server

Jupyter MCP Server

Metabase MCP server

ClickHouse MCP Server