Nadflow
Faster CI for teams that ship with coding agents. It runs the tests a change actually needs and tells you whether a failure came from your change or a flaky test.
Where Nadflow is today
On the drawing board
We're checking the problem is real and worth our time.
In the workshop
It works for us. Now we're making it work for everyone else.
Getting ready to ship
Built and tested. Finishing the last details before launch.
Live
Out in the world, and we keep improving it.
Coding agents write code faster than CI can check it. Test suites grow, runs queue up, a broken change slips into main while a flaky test blocks a good one. Nadflow is how we are keeping CI fast without giving up trust in it.
What we’re building
- Test selection. For each change, work out which tests actually need to run, from the import graph and test history. When it is not sure, it runs everything. It starts in shadow mode, so you can see what it would have skipped before relying on it.
- A verdict for every failure. Did this change break the test, is it flaky, or is it unrelated? Every verdict comes with its evidence, and “unknown” is an honest answer.
- Flaky test triage. Unreliable tests are flagged and taken out of the merge path by suggestion. Nothing is hidden automatically.
Good to know
Your source code stays in your CI: the command-line tool will be open source. We are starting with TypeScript monorepos on Turborepo, pnpm and Moon. Nadflow is a working name.