# Coder Eval

> Open-source framework to evaluate and benchmark AI coding agents (Claude Code,
> Codex, Google Antigravity/Gemini) and Claude Code skills: sandboxed runs of real
> agents against declarative YAML tasks, weighted 0.0-1.0 scoring, a
> skill_triggered activation check, A/B experiments, and CI gates.

Install: `uv tool install coder-eval` (Python 3.13+, Apache-2.0).

- Docs: https://coder-eval.com/docs
- Full docs as one file: https://coder-eval.com/llms-full.txt
- Docs llms.txt (per-page index): https://coder-eval.com/docs/llms.txt
- How it compares (vs SWE-bench, SkillsBench): https://coder-eval.com/docs/comparison
- Blog: https://coder-eval.com/blog
- GitHub: https://github.com/UiPath/coder_eval
- PyPI: https://pypi.org/project/coder-eval/