Today we are opening the waitlist for Vargus - a team of AI agents that read your code, generate the tests it should have, and run them on every change. It's the most ambitious thing Kerva has built so far, and we want a small group of teams in the door first.
The tests you keep meaning to write
Every codebase has the same shape - a few well-tested modules at the core, and a long tail of code that nobody has gotten around to. Vargus walks the tail. It reads the code, infers what it's trying to do, writes the tests, and tells you what it found - including the bugs you didn't know were there.
Agents, not autocomplete
Vargus isn't a "suggest a test" plugin. It runs as a set of cooperating agents - one to map the codebase, one to design the test plan, one to write and run, one to triage the failures. You review a pull request, not a prompt.
"Most teams treat tests as something they'll do later. Vargus is built around the assumption that 'later' is a model away." Kavan Seggie, Founder
What's in the beta
The first cohort gets Vargus running on a single repo, with automated test generation, PR-level reporting, and a daily summary of what changed and what broke. We are starting small on purpose - we want feedback from teams that are willing to dig in.
Get on the list
Spots are limited and we're letting teams in week by week. Add yourself on the Vargus product page and we'll be in touch.
