--- title: "What CAT actually is" description: "A desktop app, AI assistance, and automation — three ways to test your data, tailored to how your team works." url: "https://docs.justcat.it/what-is-cat/what-cat-actually-is/" --- # What CAT actually is CAT is a testing toolkit for data teams. You describe what should be true about your data, CAT checks whether it still is, and you find out before anyone else does. CAT is where data teams learn to test their data properly — and the toolkit they keep once they have. It comes in three parts. They are not three products and not three stages — they are three ways of doing the same work, and you use whichever ones suit how your data team works today. ![CAT Studio, AI assistance and automation shown side by side, all connecting down to a single CAT project file.](three-parts-one-project.png) *Three ways of doing the same work — not a stack, not a sequence.* ## CAT Studio A full desktop application, and where most data teams begin. - Connect your data sources, build and try out queries, write and organise your tests, run them, read the results — all in one place, visually. - It is the complete job, not a preview of it. Nothing is held back for a bigger edition. - It suits anyone who would rather work with an interface than with configuration, including people who do not write code — and it is the honest answer for a data team of one. - Installation takes a minute, from the Microsoft Store. No admin rights, no infrastructure change, no ticket. ## AI assistance Assistance across the whole loop, not only the writing of tests: working out what is worth checking, expressing it, reviewing it, running it, and making sense of what comes back. - **Working out what to check.** Describe a situation in plain language and **CAT Pilot**, built into CAT Studio, turns it into a query. It will also help you connect a data source and tell you what CAT can do about a problem you are describing. - **Writing it.** Pilot drafts the test itself, and fills in names, descriptions and tags from a query — or produces the query from the name. Nobody starts from an empty file. - **Reviewing it.** A test reviews like any other change — a colleague reads it, and so can an assistant. Nothing is hidden inside a tool. - **Running it.** From CAT Studio, the command line, or a pipeline. - **Making sense of the results.** Pilot will look at a failing test with you and help work out why, and help when a query is slower than it should be. One boundary is worth stating plainly: **the data team still decides.** What matters, what "correct" means for your business, what to fix first — none of that is delegated. Assistance removes the effort of expressing and running checks. It does not remove the judgement about which checks are worth having. Pilot works through your own AI provider account, so the choice of model — and the terms it runs under — stay yours. ## Automation and continuous integration The same tests, running without anyone watching. - Command-line tooling in three forms, so testing fits the way your team already automates: a standalone **CAT CLI** for Windows, Linux and macOS, a **PowerShell module** and a **Python module**. - Plug into Azure DevOps, GitLab, GitHub Actions, Jenkins, or in practice anything that can run a command. It is a couple of lines, and results come back in a form the platform already understands. - Run checks on every deployment, after every load, or every night — and keep the results, so you can see whether things are getting better or worse. - Light enough to sit inside an existing pipeline without anyone noticing the cost. ## One project, wherever you work The three parts are not separate worlds. A CAT project — your data sources, your tests, what you expect to be true, and where the results should go — is a single piece of plain text that you own and keep with the rest of your work. That has two consequences worth knowing: - **Nothing you write is trapped.** A test authored in CAT Studio is the same test a pipeline runs at three in the morning. Moving between the three parts is not a migration, a rewrite, or an export. - **Your testing outlives the tooling around it.** The project is readable, reviewable and versioned like any other part of your work, so it survives people leaving, stacks changing, and tools being swapped. The exact anatomy of a project — sources, expectations, output formats — is covered in the [Reference](/reference/basics/introduction/). You do not need any of it to start. ## What this is worth to your data team CAT is deliberately tailored to how differently data teams actually work, and to the fact that most of them are nothing like a textbook. - **On your own, or just starting.** CAT Studio alone is a complete, honest testing practice. Many data teams stay here permanently, because their release rhythm never demands more — and that is a finished state, not a stalled one. - **Testing more than you can comfortably write by hand.** AI assistance removes the effort of expressing checks, without changing anything about how they run or who can read them afterwards. - **Releasing often, or running pipelines on a schedule.** Automation turns testing from something somebody remembers to do into part of how you deliver. - **Growing between those.** Nothing has to be redone on the way. The tests you wrote on day one keep working, which is what makes starting with the smallest possible step a safe decision rather than a throwaway one. > **🐱:** Take only what your data team needs right now. Adding a part later costs you nothing — the tests you already wrote carry straight over — so there is no advantage in setting up automation for a release cycle you do not yet have. ## In one sentence CAT is a way to state what should be true about your data, and three ways to keep checking it: a desktop application, AI assistance to help you write the checks, and automation to run them wherever your work already happens.