Does dbt Desk connect to my warehouse or run dbt?

No. dbt Desk does not run dbt build, dbt test, dbt parse or dbt show, does not read your manifest.json or catalog.json, and never connects to a warehouse. It works only from the model file and the schema.yml entry you paste, and the model is instructed never to claim it ran anything - it says 'this test asserts', never 'this test passes', and it names the dbt command you should run when a check genuinely needs one.

What happens before I sign in?

A deterministic scan runs entirely in your browser across 61 rules. It reads the Jinja, masks SQL comments and string bodies so a select * inside a comment is not flagged, extracts the refs, sources, CTEs, config and output columns, parses the schema.yml, and reports what it found. Nothing leaves the page until you ask for a run, and the scan needs no account and costs nothing.

What are the four lanes?

Review reviews the model the way a staff analytics engineer reviews a colleague's model before it merges, across nine named checks. Contract hardens it for another team to depend on: an enforced contract with a data_type per column, an access modifier, a group and an owner, versioning and a deprecation plan, plus the dependencies.yml for any cross-project ref. Unit tests writes the dbt unit_tests YAML that pins the model's logic with every upstream input mocked. Metrics defines the semantic model and the MetricFlow metrics that sit on top of the model.

Why does a commented-out ref() still count?

Because dbt renders the whole file through Jinja before any SQL is parsed. A ref() inside a SQL comment still creates a real edge in the DAG, and a commented-out config() block still configures the model. dbt Desk flags that case explicitly, because commenting out Jinja does not disable it - only deleting it, or wrapping it in a Jinja comment, does.

Is this an official dbt Labs tool?

No. dbt Desk is a derived work built on four agent skills published by dbt Labs in the dbt-agent-skills repository: using-dbt-for-analytics-engineering, working-with-dbt-mesh, adding-dbt-unit-test and building-dbt-semantic-layer. It is not affiliated with, endorsed by, or maintained by dbt Labs. dbt is a trademark of dbt Labs, Inc.

Four things to do with one dbt model

Paste a model's .sql file and, if you have it, its schema.yml entry. dbt Desk reviews it, writes the contract another team can depend on, writes the unit tests that pin its logic, or defines the metrics that sit on top of it. A deterministic scan across 61 rules runs in your browser first — no account, no charge, and nothing leaves the page until you ask for a run.

Inspect

Govern

Guard

Measure

Review

Review the model the way a staff analytics engineer reviews a colleague's model before it merges.

Drop a .sql model or a .yml file here, or paste below. choose a file

Model SQL

no model loaded

schema.yml entry (optional, but the contract, tests and metrics lanes need it)

no schema.yml pasted

Warehouse

dbt version

Paste a dbt model to begin.

Your past runs

Runs you make while signed in are saved to your account and listed here.

Four things to do with one dbt model

Review

Prescan

Working

Result

Findings

Prescan reconciliation

Lane detail

Artifact

Your past runs