# NiceEval

> NiceEval is an agent-native eval library for AI agents and coding-agent workflows: declare what a good result looks like with adapters, experiments, and evals, then run it against any agent, model, or sandboxed coding-agent session.

- [Overview](https://niceeval.com/): Product overview, audience modes (human / agent setup), and runnable eval examples.
- [Blog](https://niceeval.com/blog): Product and engineering posts about building evals with NiceEval.
- [GitHub](https://github.com/CorrectRoadH/niceeval): Source code, issues, and releases.
- [INIT.md](https://niceeval.com/INIT.md): Step-by-step instructions for an AI coding agent to integrate NiceEval into a host repository.

## Docs

- [Full documentation index](https://niceeval.com/docs/llms.txt): Concepts, guides, and reference pages for adapters, experiments, evals, scoring, sandboxes, and the CLI.
- [Quickstart](https://niceeval.com/docs/quickstart): 5-minute path to a first running eval.

## 中文

- [产品首页](https://niceeval.com/zh)
- [博客](https://niceeval.com/zh/blog)
- [中文文档](https://niceeval.com/docs/zh/tutorials/quickstart)