agent-browser

Why we included this project

For agent projects that have to drive a real browser, this CLI keeps the loop simple: you grab a snapshot of the page's accessibility tree, get stable refs for elements, then issue click, fill, type, or text commands against those refs. Because it's a native Rust binary with a flat command surface, you can call it from any language without pulling in a browser automation framework. It also handles the boring parts: downloading a pinned Chrome build, spotting existing Chrome, Brave, Playwright, or Puppeteer installs, and dealing with tabs, forms, uploads, and auth. That combination of a predictable ref-based workflow and low setup friction makes it a practical pick for agentic automation that needs scriptable, reliable browser control.