Brow Use

README excerpt

# brow-use

**Browser automation for Claude Code.** Drive a real browser from plain English —
get the answer back, or generate the Playwright tests.

brow-use is a Claude Code plugin: an MCP server plus a Chrome extension that give
an agent tools to navigate, perceive, and act on live web applications. The agent
reads the **accessibility tree** at every step — the same semantic view screen
readers use — so it sees what's actually on the page instead of guessing at
selectors from a screenshot.

> Status: spike / `v0.1.0`. APIs and artifact layouts are still moving.

---

## What you can do with it

Four capabilities, all powered by the same browser-driving loop. Each is usable on
its own; together they compose.

| | Capability |
|---|---|
| **Build test infrastructure** | Generate typed Playwright Page Object classes and reusable workflow functions from real, recorded interactions with a live app — not guessed from screenshots. |
| **Fetch data in natural language** | Describe what you need ("export the active subscriber list as CSV") and get it in the format you asked for. Works behind authentication without sharing credentials with the model. |
| **Perform actions in natural language** | Describe the outcome ("file this expense report"); the agent carries it out and leaves a forensic record — a Playwright trace plus a reasoning log. |
| **Document an application** | Produce end-user feature docs for a whole web app: one page per cluster of screens, embedded screenshots, a na