How do I get started with CAT?

Start with Claude Code: install the CAT CLI, add the CAT skills, and ask for the checks you need. Prefer an app? Install CAT Studio in a minute and let CAT Pilot draft your first check.

What can my data team test with CAT?

If you can describe what should be true about your data, you can test it: data quality, consistency across systems, business rules, the meaning behind every number — all the way to the report.

Which data platforms does CAT work with?

CAT is close to technology agnostic: if you can query it, CAT can test it. From SQL Server, Snowflake, Databricks or Power BI to the Excel and CSV files the business still works in.

What does “AI-native” mean for CAT?

Agents do the work. Our documentation is ready for agents, the CAT skills load it into Claude Code or any coding agent, and with the CAT CLI your agent writes, runs and maintains your checks — on your desktop or from your pipelines.

Our company restricts AI. Can we still use CAT?

Yes. CAT Studio, the CLI and your pipeline all work without AI. Where AI is allowed, your IT team decides: CAT Pilot can be switched off for the whole organisation, or limited to approved providers, through Group Policy or Intune.

Does our data leave our network?

No. CAT runs on your own machines, against your own data sources, and your tests, results and data stay with you. Only when you bring AI into the work — a public model or a private endpoint — does your company's AI policy apply, as it would anywhere else in your organisation.

We built our own testing scripts. Do we have to start over?

No. The checks you already have hold real knowledge about your data. Hand them to your coding agent and it moves them into CAT for you — and you drop the upkeep of the framework around them.

We already use DQX. Why would we need CAT?

Keep it — checks inside one pipeline are useful, but far from enough. CAT looks across the whole data solution, every layer from the first load to the last report, and shows you what a change really did.

Is CAT for a one-person data team or a large company?

Both. A data team is a role, not a headcount: if people rely on the data you produce, you are the data team. One person starts in CAT Studio or Claude Code; a dozen data teams share the same toolkit and project format, so what one builds, the next can reuse.

Is CAT a data governance or observability platform?

Neither. We have never seen one make a data team's daily work easier — and that daily work is exactly what CAT is for. CAT is a toolkit built for the everyday needs of data teams, straight from the trenches.

How much does CAT cost?

CAT starts free, so your data team can prove the value before anyone signs anything.

Claude Code Compatible

Make your data team get quality assurance right.

CAT is an AI-native data quality toolkit,
so you know whether to ship, or not yet.
Every time.

Download CAT Studio Read documentation

CAT CLI available for Windows, macOS and Linux

CAT Studio

Not runCustomers match the CRM

Not runOrders match the shop

Not runYesterday’s payments all arrived

Not runInvoices match SAP

Not runOrder totals add up to their items

Not runPaid invoices are closed

Not runEach invoice belongs to a real order

Not runNo customer is missing a sales region

Not runNothing overdue is already paid

Not runPrices follow the price list

Not runVAT rates are valid for the customer’s country

Not runOpen balances agree with SAP

Not runLast month stays closed

Passed0

Failed0

Total13

Project13

Pass rate—

Duration—

Finished—

CAT Pilot

AnthropicClaude Sonnet 5.5

Trusted by data teams at

Why data teams stay with CAT
to keep their delivery routine rock solid

Work AI-native, for real
Write and maintain your checks in plain words, or let your coding agent do it — the way data teams work today, not someday.
Test with shared practices
No blank page: your data team writes, runs and reviews checks the same way — and nobody is lost anymore.

Use a toolkit, not a platform
Run your checks where you already work — on your machine or in your pipeline — with no platform to roll out, administer or feed.
Release anytime, with confidence
Verify your data as a reliable routine, release without fear — and earn the trust of the business.

Unit tests are
not enough.

See what a change did to the whole data solution, not just to what you touched: every layer downstream, every report finance opens on Monday.

What your data team can take on

Behind every great data team
is quality assurance done right.

brew install justcatit/cat/catcli

Or read the documentation

Revenue in the Sales report must match finance, month by month. Write a check for it.

Drafted. It reads like your rule:

Check:Revenue in the Sales report matches financeCompare:Sales report · Total Revenue, by monthwithFinance · invoices, by month

Expect:every month the same

PASS12 months compared · done at 1:03 AM

Finance now books refunds separately.

Updated one line — the finance side now subtracts refunds. Same rule, same name.

Expect:every month the same

PASS12 months compared · done at 1:06 AM

Revenue in the Sales report must match finance, month by month. Write a check for it.

Drafted. It reads like your rule:

Check:Revenue in the Sales report matches financeCompare:Sales report · Total Revenue, by monthwithFinance · invoices, by month

Expect:every month the same

PASS12 months compared · done at 1:03 AM

Finance now books refunds separately.

Updated one line — the finance side now subtracts refunds. Same rule, same name.

Expect:every month the same

PASS12 months compared · done at 1:06 AM

Can we ship the new customer dimension?

I added two checks next to your change and ran everything it could reach.

Your change · dim_customer2/2 passed

Staging12 / 12 passed

Core34 / 34 passed

Data marts21 / 21 passed

Reports9 / 9 passed

Checks reviewed with the change by a colleague
Release pipeline runs the same checks — 78 passed

PASS · ship it78 passed · done at 9:03 PM

2 checks failed downstream of the customer dimension release — nobody had to go and look. One of them is the check you wrote last week: Revenue in the Sales report matches finance.

Data mart · sales_by_regionFAIL

Report · Sales by region (Power BI)FAIL

Orders in core, missing in the mart:

ORD-118204 ORD-118205 ORD-118211 …

Sales report revenue lower than finance in 3 regions

FAIL · you see it firstDaily · done at 3:13 AM

Why? Everything passed before the release.

Production has 214 new customers without a region — the test data had none. The mart's join drops their orders.

Fix proposed: a left join with an "Unknown" region. Pull request ready for your review.

REVENUEPASS

ORDERSPASS

CUSTOMERSPASS

STOCK LEVELSPASS

PRODUCT FINANCIALSPASS

Overnight run finished 06:42
Yesterday's fix is in; sales_by_region and the Sales report pass again
Your agent went through the night's runs — nothing failed
One to watch: stock levels arrived 40 minutes later than usual, still within the window
The business sees the same answer on its data quality report

PASS · nothing needs you this morningDaily · done at 5:09 AM

Other ways to use CAT:

macOS Linux Claude Code Terminal GitHub Copilot Cursor Codex Python

Wherever your data team works

Three ways into the same work. Pick the one that fits your data team today, and add another anytime, without rewriting a single test.

Claude Code running CAT data quality tests in the terminal

AI-native testing

Work with your coding agent

Claude Code or any coding agent writes, runs and maintains your checks — you review them like any other change. A CAT test is plain text in your repository, and the CAT CLI runs headless on Windows, Linux and macOS.

Get the CAT skills CLI docs

Windows · Linux · macOS

CAT Pilot in CAT Studio suggesting the next data issue to fix

Plain-words testing

Work in CAT Studio

A desktop app where the business can be part of data quality too. Say in plain words what should be true, and CAT Pilot — your AI assistant — turns your words into checks you can run and maintain — while building the context your agents work from. No code needed.

Get CAT Studio Documentation

Windows · Microsoft Store · no admin rights

CAT tests running in an Azure DevOps pipeline

Continuous testing

Let it run on its own

Whatever you wrote with your agent or in CAT Studio runs in your pipeline unchanged — on every deployment, after every load, every night. Azure DevOps, GitLab, GitHub Actions, Jenkins or anything that runs a command — with every result ready to be picked up by your agents.

Pipeline guides

Azure DevOps · GitHub Actions · GitLab · Jenkins · any CI

One `project file` underneath all three — and it is yours.

Plain text in your repository: readable, reviewable and versioned like the rest of your work — and still yours, whatever tools come and go.

Make your data team get quality assurance right.

Why data teams stay with CAT
to keep their delivery routine rock solid

Work AI-native, for real

Test with shared practices

Use a toolkit, not a platform