pi-agent-browser

Browser automation tool for pi. Gives the LLM a browser tool that drives a real browser via agent-browser.

Install

pi install npm:pi-agent-browser

Or try it without installing:

pi -e npm:pi-agent-browser

What it does

Registers a browser tool that the LLM can call to:

• Navigate — open <url>
• Inspect — snapshot -i (returns interactive elements with @ref handles)
• Interact — click @e1, fill @e2 "search query", press Enter, scroll down
• Read — get text, get title, get url, get text @e3
• Screenshot — returns the image inline so the LLM can see the page
• Clean up — close (also auto-closes on session shutdown)

Features

| Feature | Details | |---|---| | Inline screenshots | Screenshots are returned as base64 images — the LLM can describe what it sees | | Output truncation | Large snapshot output is truncated to fit context windows, with full output saved to a temp file | | Auto-install | If agent-browser isn't installed, prompts to install it (npm + Chromium download) | | Session cleanup | Browser is automatically closed on pi session shutdown — no orphaned Chromium processes | | TUI rendering | Compact display: shows command inline, element counts for snapshots, screenshot paths |

Example

You: Open hacker news and tell me the top 3 stories

browser open https://news.ycombinator.com
browser snapshot -i
browser close

The top 3 stories on Hacker News right now are:
1. ...
2. ...
3. ...

Requirements

• agent-browser — installed automatically on first use, or manually:

  npm install -g agent-browser
  agent-browser install  # downloads Chromium

• A vision-capable model (for screenshot descriptions): Claude Sonnet/Opus, GPT-4o, Gemini Pro, etc.

License

MIT