# unotest web — AI-native E2E testing for web apps

> For the full site index, see [llms.txt](/llms.txt)

unotest web is AI-native end-to-end testing for web apps. Your AI agent
drives your real app over MCP and writes a clean, reviewable `.js` test in your
repo — readable on top, precise underneath. You review and commit.

Install: `npx @unotest/web init` — package `@unotest/web`.

## How it works

1. **Describe a flow in plain English.** Tell your agent what to test.
2. **The agent writes and runs it.** It drives your real app over MCP and emits
   a clean `.js` scenario, then runs it through the sandboxed interpreter.
3. **It breaks, it fixes.** On failure the run pauses mid-scenario; the agent
   inspects runtime state, patches the step, and resumes — no full restart.

## What you get

- **A Playwright-vocabulary DSL** on a sandboxed engine. Every step carries a
  plain-English intent; inside it sit the exact DSL calls.
- **MCP-native.** The agent has a full MCP toolset to explore, run, debug and
  author tests against your live app.
- **A local viewer** to watch runs, inspect steps and artifacts.
- **Plain `.js` in your repo.** Git, code review, CI — no proprietary format.
- **No silent fixes.** Every change the agent makes is a visible diff you review.
- **Runs locally and safe.** The DSL executes in a vendored AST interpreter —
  no `require`, `fetch` or filesystem access.

## Works with

Claude Code, Cursor, Codex.

## Documentation

https://docs.unotest.com