# kensa

> Kensa turns agent traces into pytest evals that run in CI.

## Docs

- [Assertions](https://kensa.sh/docs/assertions.md): Deterministic trace and output assertions. Fast, free, and they gate the judge.
- [Cases](https://kensa.sh/docs/cases.md): Define agent inputs and trials with kensa_case.
- [Changelog](https://kensa.sh/docs/changelog.md): What's new in Kensa.
- [CI](https://kensa.sh/docs/ci.md): Run Kensa evals in continuous integration.
- [CLI reference](https://kensa.sh/docs/cli.md): Every kensa command and its options.
- [Concepts](https://kensa.sh/docs/concepts.md): The mental model behind Kensa. Cases, traces, assertions, judges, trials, and eval ideas.
- [Introduction](https://kensa.sh/docs/introduction.md): Kensa turns agent traces into pytest evals that run in CI.
- [Judge](https://kensa.sh/docs/judge.md): LLM-as-judge for semantic criteria that resist deterministic checks.
- [Pytest plugin](https://kensa.sh/docs/pytest.md): The marker, fixtures, and harness that make Kensa evals plain pytest tests.
- [Quickstart](https://kensa.sh/docs/quickstart.md): Add Kensa, wire the harness, and land your first eval.
- [Skill](https://kensa.sh/docs/skills.md): The kensa-evals skill drives the eval lifecycle from your coding agent.
- [Tracing & imports](https://kensa.sh/docs/tracing.md): Bring real agent behavior into Kensa from Langfuse, a trace export, or local capture.

## OpenAPI Specs

- [openapi](https://kensa.sh/docs/api-reference/openapi.json)