Skip to content

AWLSEN/orb-browser

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

121 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

orb-browser

Give your AI agent a browser.

A real Chrome browser in the cloud that your AI coding agent can control. It scrolls, clicks, types, screenshots — everything a human can do. When idle, it freezes to disk ($0). Wakes in 500ms, still logged in.

Install the Claude Code Plugin

pip install orb-browser

In Claude Code:

/plugin marketplace add AWLSEN/orb-browser-marketplace
/plugin install orb-browser@orb-browser

Then just talk to your agent:

"Go to Hacker News and tell me the top 5 stories"
"Watch my Twitter for AI posts every 2 hours and email me a summary"
"Screenshot apple.com and describe the layout"
"Fill out the contact form on example.com"

Your agent handles the rest — deploys the browser, browses, reads, acts, sleeps.

What Can It Do?

  • Browse any website — navigate, scroll, click, type, fill forms
  • Read pages — extract text, take screenshots, run JavaScript
  • Log into sites — grabs cookies from your Chrome automatically
  • Monitor on schedule — wake every N hours, browse, email you, sleep
  • Watch it live — opens a real-time browser view so you see everything

How It Works

  1. Your AI agent (Claude Code, Codex) gets a cloud browser via the orb-browser CLI
  2. The browser runs real Chrome on Orb Cloud
  3. Agent browses: orb-browser go, orb-browser text, orb-browser click
  4. When done: orb-browser sleep — browser checkpoints to NVMe, $0/hr
  5. Next run: orb-browser wake — 500ms, everything restored, still logged in

Login Flow

For sites that need login (Twitter, Gmail, etc.), the agent grabs cookies from your local Chrome automatically:

import browser_cookie3
cookies = browser_cookie3.chrome(domain_name='.x.com')

Injects them into the cloud browser. Checkpoints. Next wake = still logged in.

CLI Reference

orb-browser deploy              # Create cloud browser
orb-browser go <url>            # Navigate
orb-browser text                # Read page text
orb-browser screenshot [path]   # Take screenshot
orb-browser click <x> <y>      # Click coordinates
orb-browser click <selector>    # Click CSS selector
orb-browser type <text>         # Type text
orb-browser fill <sel> <val>    # Fill input
orb-browser scroll [down|up]    # Scroll
orb-browser eval <js>           # Run JavaScript
orb-browser cookies             # Get/set cookies
orb-browser sleep               # Checkpoint ($0)
orb-browser wake                # Restore (500ms)
orb-browser live                # Open live view
orb-browser destroy             # Delete VM

Python SDK

from orb_browser import OrbBrowser

orb = OrbBrowser(api_key="orb_...")
orb.deploy()

orb.navigate("https://example.com")
print(orb.text())
orb.screenshot("page.jpg")

orb.sleep()   # $0
orb.wake()    # 500ms, still on example.com
orb.destroy()

Cost

State Cost
Running ~$0.03/hr
Sleeping $0/hr
1,000 browsers, 90% sleeping ~$50/month

License

MIT

About

Give your AI agent a browser. Claude Code + Codex plugin.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors