Find what your agent’s
evals miss
Your agent changed. Your evals didn't. Our AI engineer reads your traces next to your evals and shows you what they let through.
curl -fsSL https://kitaru.ai/install | bash
Eval therapy
You answer a few questions. Good evals fall out.
You don't write evals from scratch. Your AI engineer asks only what it can't work out alone.
- 01
Connect your repos
Connect GitHub. It reads your repos and finds your agent, your evals and your policies. Then add your tracing keys.
- 02
A short interview
It asks about what it found. Talk it through or type.
- 03
Hear what your evals miss
Right after, it shows what your evals missed, in the order you care about.
- 04
Fifteen minutes a week
Ten quick questions on three real sessions.
How it fits
An intelligence layer on the stack you already run
Nothing to rewrite, no new framework. It reads your repos and traces through keys you can revoke.
- 01Reads a traceSession 2207 · refundRefunded without finding the order
- 02Finds a gapAnswers before looking up the order58 sessions like it, no eval checks it
- 03Writes a checkorder_looked_up_firstAgrees with your reviewers 57 of 58 times
- 04Tests the fixPrompt fix, replayed via OpenAI Agents SDK58 of 58 now pass
Your traces
Your evals
- LLM judges
- Scores in your tracer
Your AI engineer
Intelligence layer
Your agent
Trusted to run production AI, from pipelines to agents, at
Open source underneath
Orchestrate AI on the infra you choose
Run pipelines and agents in Python on the orchestrator and cloud you already have, with every run, artifact, and model version recorded.
Test every agent update on real traces
Import the traces you already have, review sessions, turn your verdicts into evaluators, and replay any change before it ships.
Your AI engineer works on top of both. Use them on their own, free, on your own machines.
Bring one agent
and a week of traces
Connect your repo and your tracing keys. By the end of the interview you know what your evals miss. 14 days free, no credit card.