- Category
- Developer Tools
- Rank
- No. 1823Tools index
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Model & Agent Evaluation · Deployment & Operations
- GitHub
- 4 stars
- Latest release
- v0.1.3
- Date
About
Apgar is a release health checker for AI agents. For the first hours of each new release it compares agent behavior from production traces against the previous release on metrics such as tool errors, retries, loops, answer length, cost and latency. It sends a Slack message or webhook naming the metric that moved and where the change is concentrated. It uses no LLM judge, so it can run on all traffic, but it cannot detect wrong answers that look normal.
Why it made the leaderboard
It flags when a harness change makes an agent worse in production, cheaply and on full traffic, catching regressions that sampled LLM-judge evals miss.
Tags
observabilityai-agentsopentelemetrylangfusereleasesregression
Tech Stack
PythonDocker
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.