dbt Labs

Fetches dbt's official docs in markdown and answers questions about dbt Cloud, Core, and the Semantic Layer without making you tab over or guess at syntax.

Reads your dbt manifest and generates a clean Mermaid diagram showing which models feed into which — helps spot circular refs and data lineage gaps at a glance.

Generates test YAML that mocks upstream inputs and validates your model output. Helps you catch logic bugs before they hit production.

Reads security or quality audit reports from skills.sh or Tessl and recommends specific remediations, prioritized by risk and ease of fix.

Scaffolds dbt models, writes SQL transformations with proper ref() and source() calls, adds tests, and validates them — all with knowledge of your warehouse schema and dbt conventions.

Reads the run log, checks git history, queries the Admin API, and traces the actual data issue — not just the error message. Returns a root cause and next step.

Migrates your dbt models from one warehouse to another (Snowflake to Databricks, etc.) by identifying SQL dialect incompatibilities and rewriting them in real time.

Parses your intent ("test the new fact table"), finds the right dbt binary, and executes the formatted command. Handles model selection, test runs, and builds.

Asks your data warehouse for revenue, churn, conversion rate, or any other metric in plain English. Uses your dbt Semantic Layer for consistent definitions across the org.

Generates the MCP config file, walks you through auth setup, and tests that Claude can actually query your dbt project—no manual JSON editing required.

Writes and validates dbt Semantic Layer YAML — semantic models, metrics, dimensions, entities, and time spines. Handles MetricFlow config and all metric types without boilerplate errors.

Runs dbt-autofix, then sorts remaining errors into buckets: auto-fixable, guided fixes, needs your input, or blocked. Points you to the right fix for each.

Configures dbt Mesh contracts, access modifiers, groups, and cross-project dependencies so multiple teams can share models safely without breaking each other's pipelines.