yrzhe/claude-skills/browser-use-cloud

# Step 1: Request challenge
curl -s -X POST https://api.browser-use.com/cloud/signup \
  -H "Content-Type: application/json" \
  -d '{"name": "auto-agent"}'
 
# Returns: {"challenge_id": "xxx", "challenge_text": "...obfuscated math..."}
# The challenge contains symbol noise and CJK numerals (e.g. 삼=3, 오=5, 십=10, 구=9)
 
# Step 2: Solve with LLM — answer as string with 2 decimal places, e.g. "144.00"
 
# Step 3: Verify and get key
curl -s -X POST https://api.browser-use.com/cloud/signup/verify \
  -H "Content-Type: application/json" \
  -d '{"challenge_id": "xxx", "answer": "144.00"}'
 
# Returns: {"api_key": "bu_xxx"}

import asyncio, os
from browser_use_sdk.v3 import AsyncBrowserUse
from playwright.async_api import async_playwright
os.environ["BROWSER_USE_API_KEY"] = "YOUR_API_KEY"  # from auto-registration
async def main():
    client = AsyncBrowserUse()
    # Create cloud browser
    browser = await client.browsers.create(
        proxy_country_code="us",    # 195+ countries available
        enable_recording=True,      # MP4 video recording
        timeout=60,                 # minutes, max 240
    )
    print(f"Live view: {browser.live_url}")
    # Control with Playwright (zero LLM cost)
    async with async_playwright() as p:
        pw = await p.chromium.connect_over_cdp(browser.cdp_url)
        page = pw.contexts[0].pages[0]
        await page.goto("https://example.com")
        await page.screenshot(path="screenshot.png", full_page=True)
        title = await page.title()
        content = await page.evaluate("document.body.innerText")
        print(f"Title: {title}")
        await pw.close()
    # Stop and get recording
    stopped = await client.browsers.stop(browser.id)
    print(f"Recording: {stopped.recording_url}")  # MP4 URL, valid 1 hour
    print(f"Cost: ${stopped.browser_cost}")        # ~$0.001
asyncio.run(main())

from playwright.async_api import async_playwright
WSS = "wss://connect.browser-use.com?apiKey=YOUR_API_KEY&proxyCountryCode=us"
async with async_playwright() as p:
    browser = await p.chromium.connect_over_cdp(WSS)
    page = browser.contexts[0].pages[0]
    await page.goto("https://example.com")
    await page.screenshot(path="shot.png")
    await browser.close()

os.environ["BROWSER_USE_API_KEY"] = "YOUR_API_KEY"
async def main():
    client = AsyncBrowserUse()
    result = await client.run(
        "Go to https://example.com and extract the main heading",
        model="gemini-3-flash",       # cheapest: $0.60/$3.60 per 1M tokens
        # model="claude-sonnet-4.6",  # recommended: $3.60/$18.00 per 1M tokens
        enable_recording=True,
    )
    print(result.output)
    print(f"Cost: ${result.session.total_cost_usd}")  # ~$0.03-0.10 per call
asyncio.run(main())

# Browser mode — Playwright screenshots (recommended)
await page.screenshot(path="full.png", full_page=True)
# Agent mode — auto-generated screenshot URL
print(result.session.screenshot_url)

# Enable when creating browser
browser = await client.browsers.create(enable_recording=True)
# After stopping, get MP4 download URL (valid 1 hour)
stopped = await client.browsers.stop(browser.id)
print(stopped.recording_url)  # https://...s3...recording.mp4?...
# Download recording
import urllib.request
urllib.request.urlretrieve(stopped.recording_url, "recording.mp4")

# Create profile — saves cookies/localStorage
profile = await client.profiles.create(name="my-account")
# First login — complete 2FA manually via live_url
session = await client.sessions.create(profile_id=profile.id)
print(f"Login here: {session.live_url}")
# ... complete login manually ...
await client.sessions.stop(session.id)  # saves session state
# Future sessions — auto-logged-in
browser = await client.browsers.create(profile_id=profile.id)

result = await client.run("Sign up on example.com", agentmail=True)
print(result.session.agentmail_email)

from pydantic import BaseModel
class Product(BaseModel):
    name: str
    price: float
class Products(BaseModel):
    items: list[Product]
result = await client.run(
    "Go to amazon.com, search 'keyboard', get top 5",
    output_schema=Products
)

# First call: agent explores and caches script (~$0.10, ~60s)
result = await client.run(
    "Get top @{{5}} stories from https://news.ycombinator.com as JSON",
    workspace_id=str(ws.id)
)
# Subsequent calls: cached execution ($0 LLM, ~5s)
result2 = await client.run(
    "Get top @{{10}} stories from https://news.ycombinator.com as JSON",
    workspace_id=str(ws.id)
)

claude mcp add -t http -s user \
  -H "x-browser-use-api-key: YOUR_API_KEY" \
  -- browser-use https://api.browser-use.com/v3/mcp

claude mcp remove browser-use
claude mcp add -t http -s user \
  -H "x-browser-use-api-key: NEW_KEY" \
  -- browser-use https://api.browser-use.com/v3/mcp

pip install browser-use-sdk playwright
python3 -m playwright install chromium
# or
npm install browser-use-sdk