Apgar
AI Dev Platforms · MLOps & Evaluation 100% Free Open Source

Apgar

Deterministic agent regression testing suite catching tool-calling defects without expensive LLM judges.

User Satisfaction 100%
19 Community Reviews

Free Tier & Quota Rules

Open-source test harness catching agentic behavioral regressions in CI without relying on costly LLM judges

Overview & Core Features

Apgar redefines agentic software reliability. Instead of spending thousands on nondeterministic LLM-as-a-judge pipelines that fluctuate unpredictably, Apgar records deterministic trajectory graphs, argument schemas, and state transitions, catching behavioral regressions in milliseconds during standard continuous integration runs.

Pros & Limitations

Strengths

  • ✓ 100% deterministic and reproducible test executions removing LLM judge temperature noise
  • ✓ Zero evaluation API bills enabling fast tests on every single git commit
  • ✓ Pinpoints exact trajectory steps and parameter mismatches without opaque guessing

Considerations

  • ✕ Requires engineers to curate initial baseline trajectory recordings and contracts
  • ✕ Unsuited for purely subjective open-ended creative storytelling evaluations

Ready to try Apgar?

Explore features directly on their official platform.

Open Apgar ↗

Similar Alternatives

💬 Discussion

GitHub Powered