Look up before you make up.

Your agent is guessing what your data means.

Bring the schema your stack already produces — a dbt manifest, a JSON Schema, a warehouse extract. No passwords. Give the columns that matter a definition a person can edit and an agent can read.

Connect dbtSee all connectors

Three steps

How it works

  1. 1Step 1

    Bring a file.

    dbt parse writes manifest.json. That's the whole import.

  2. 2Step 2

    Fix the meaning in a grid.

    Permitted values, definitions, owners, the mistakes an agent keeps making — in a grid the finance lead can edit, not a YAML file.

  3. 3Step 3

    Point your agent at it.

    A read-only endpoint your assistant, Claude, Cursor or internal agent reads before it answers.

dbt logodbtJSON Schema logoJSON SchemaSnowflake logoSnowflakeSalesforce logoSalesforceGoogle BigQuery logoGoogle BigQueryDatabricks logoDatabricksConfluent logoConfluentREDCap logoREDCap

16 platforms · 10 formats · labelled for what you can do today →

Stop your agent guessing what `status` means

An LLM writing SQL against your warehouse has three bad options: guess from the column name, read the SQL, or ask a human. Give it a fourth.

  • Import the manifest; each accepted_values list becomes a governed vocabulary.
  • Fill the three columns agents get wrong, meaningNote, commonMistake, doNotUseFor, on the 20 columns that matter.
  • Connect your assistant to the read-only endpoint and replay a question it used to get wrong.
Start this recipe

Who this is for

You run dbt and you are the person everyone asks.

Two `accepted_values` tests disagree and both pass. Someone wants “active customer” changed and it’s been three weeks. This is where that ends.

You pointed an assistant at the warehouse and it made something up.

It inferred what `status` means from the column name. It was confident. It was wrong. Give it something to look up.

Pricing

A library of connectors.
Simple pricing per builder, per team, or for the whole enterprise

Solo work. Team evaluation.

Builder

$49/ month

  • 1 builder seat · up to 3 projects
  • Up to 100,000 nodes per model
  • Unlimited view-only collaborators
Most Popular

Where teams build in production.

Team

$990/ month

  • 10 builder seats included · $99/month per additional seat · up to 10 projects
  • Up to 1,000,000 nodes per model
  • Unlimited view-only collaborators

Founding teams: $990 $490 / month for 12 months · then $990 / month · first 10 teams · ends 30 September · 10 of 10 remaining. See the founding offer.

No limits. Governed at scale.

Enterprise

from $4,000/ month

  • Unlimited projects, users & schema scale
  • SSO · SLA · private vaults
  • A named contact
See plans