brow

Why brow?

Most browser automation tools are designed for test suites: verbose configuration, assertion libraries, test runners. brow is designed for agents: terse commands, structured output, minimal surface area, persistent profiles.

• Session-based: each session is a real Chromium profile — log in once, reuse forever
• Structured output: snapshots return accessibility trees, not raw HTML — easy for LLMs to parse
• API-first: all commands go through a local HTTP daemon — call it from any language
• Action recording: every interaction is logged and can be replayed or exported as a script

Install

pip

pip install brow-cli
playwright install chromium

Homebrew

brew tap detrin/tap
brew install brow

Agent skill

npx -y skills add detrin/brow

Quick start

brow session new --headed           # → 1
brow navigate -s 1 "https://news.ycombinator.com"
brow snapshot -s 1
brow click -s 1 "text=Ask HN"
brow session delete 1

Architecture

Claude Code / your script
        │
        ▼  CLI (brow)
   BrowClient (httpx)
        │
        ▼  HTTP  localhost:19987
   FastAPI daemon  (auto-starts on first command)
        │
        ▼  Playwright
   Chromium browser

The daemon starts automatically the first time you run a command that needs it. It keeps sessions alive across multiple CLI invocations — you don’t need to pass credentials or re-login between commands.