Connect dbt to AI: Let Claude and Cursor Read Models, Tests, and Lineage

Why Connect dbt to AI?

dbt is where data teams define transformations, run tests, and track lineage. When you need to check a model's status, verify a test passed, or trace upstream dependencies mid-coding, you open dbt Cloud, navigate to the right job or model, read the result, switch back.

With an MCP server pointed at dbt Cloud, AI agents can query your project directly. Check run status, read model definitions, browse lineage graphs without leaving your terminal.

What AI Agents Can Do With dbt Access

Once connected, Claude, Cursor, or Codex can:

• Check runs — "Did the nightly dbt run succeed?"
• Read models — "What does the dim_customers model depend on?"
• Browse lineage — "What models feed into the revenue_metrics mart?"
• Check tests — "Which dbt tests failed in the last run?"
• Query sources — "What sources are defined in the staging layer?"
• Read docs — "What's the description of the orders model?"

How It Works

1. Open DataFaucet and enter your dbt Cloud URL
2. Browse dbt Cloud — click through jobs, models, lineage, tests
3. DataFaucet captures the read operations as MCP tools
4. Deploy — your AI agents now have structured access to query dbt

You're browsing the dbt Cloud web interface, not sharing API tokens. DataFaucet captures HTTP interactions from the web UI. Your dbt credentials stay in your browser session.

Example Queries

"Did the production dbt job succeed this morning?"
"What models are in the marts/finance directory?"
"What tests are defined on the orders model?"
"What's the lineage for fct_monthly_revenue?"
"Which sources are stale?"

When This Saves Time

• Pipeline debugging — check run status without opening dbt Cloud
• Code review — verify model dependencies before approving PRs
• Development — check if upstream models have the columns you need
• Incident response — quickly find which tests failed and why
• Documentation — query model descriptions during discussions

Security Model

DataFaucet connects through the dbt Cloud web interface. Access is scoped to whatever your dbt Cloud permissions allow. If you only have access to certain projects or environments, the MCP server has the same constraints. No API tokens stored separately.

Getting Started

Open dbt Cloud in your browser, copy the project URL. Paste it into DataFaucet, browse through jobs, models, and lineage. Deploy. Your agents can now query your transformation layer from your terminal.

Create your dbt MCP server →