Ancora finds the bugs that only show up in production.
Point it at a repo. Ancora deploys the app in containers with seeded data, generates a test suite that drives it from the API down to the database, and delivers every bug as a runnable reproduction.
$ pytest suites/concurrency/test_saves.py test_save_roundtrip PASSED test_save_during_slow_external_call FAILED AssertionError: note body lost after concurrent save expected: the user's edit got: "" reproduction written to repros/save_race.py rerun anytime: pytest repros/save_race.py
A save lands while a slow external call is in flight and the change silently disappears. This bug exists only in the assembled system, so no unit test can reach it.
Run on six open source projects so far: 18 confirmed bugs, none previously reported, each with a runnable reproduction.
How it works
-
1
Point it at a repo.
Ancora figures out how to deploy the app, stands it up in containers, and seeds it with data.
-
2
It generates a test suite.
The agent infers what the app is supposed to do, checks in with you on what to test, then drives the running system with real state, fault and delay injection, and concurrent load.
-
3
Every bug arrives as a runnable reproduction.
Everything Ancora produces is pytest, so failures rerun on demand, with or without an AI in the loop, and hand straight off to manual or agentic debugging.
Why these bugs get missed
The hardest bugs live between components: a race that loses user data, an integration that writes a blank file over saved work, a partial failure that leaves state half committed. Unit tests cannot reach them and browser tests cannot see them.
Catching them takes a live environment running the assembled system. That environment has always been expensive to build, so most teams skip it and meet these bugs in production. Ancora generates it from the repo.
Founders
Gene Hong
Builds data and agentic AI products, where evals and verifying that systems actually work are the everyday bottleneck.
Dan Qian
PhD researcher in distributed systems, working on AI agents that build and verify them. Previously built data systems at AWS and Bloomberg.
Miles Liu
Built ML and systems software for surgical robotics at Intuitive Surgical, where a software failure breaks physical hardware.
Launching publicly soon.
We are onboarding early users now. If you want Ancora on your codebase, get in touch.