For data teams
From a question
to a merged
pipeline.
Someone asks a question the warehouse can’t answer yet. Take it to Jitto Spec to write a spec with checks. Jitto Build builds against them and gives you a change to review and merge.
Bring one question your team can’t answer quickly today.
| Step | What happened | Result |
|---|---|---|
| Ask | Which sales channel brought in the most last quarter? | No data yet |
| Spec | 3 requirements, each with a check | Approved |
| Build | The check failed, then passed | Green |
| Merge | A person reviewed the change | Merged |
How it works
Four steps.
A person signs off.
Ask
Ask Jitto Analyze in plain English. Every answer shows the query it ran and the rows it got back. When Jitto Analyze can’t read the data, it says so instead of guessing.
Spec
Jitto Spec asks the person who needs the answer what they mean, and writes each requirement with a check. The owner approves it.
Build
Jitto Build works against the checks in a real terminal. Anyone on your team can take over. The check fails first and passes after.
Merge
A person reviews the change and merges it. Approval is refused while a check is red.
What you keep
Your code.
Checks you can read.
Models as code in your git
Jitto Build writes the models as code, with a test on every model, in your own repository.
Checks apart from the agent
The checks compare row counts, totals and keys. The agent doesn’t mark its own work.
A record for every step
Each step writes what it read, what it made and whether it passed.
An MCP endpoint for your own agents
Jitto Analyze has an MCP endpoint, so your own agents can ask it questions.
Questions
Before you start.
Does Jitto merge its own work?
No. A person reviews each change and merges it.
Can I step in while it builds?
Yes. Take over the terminal session and carry on from where it stopped.
Does it change our source systems?
No. It reads them and copies data into the warehouse.
Is the example real data?
No. The question and the steps above are invented.