Browser Use

Browser Use is a browser automation sdk that uses screenshots to capture the state of the browser and actions to simulate user interactions. This chapter will introduce how you can easily use browser-use to execute agent tasks on the Web with simple calls.

Get Scrapeless API Key

Go over the Dashboard’s Settings tab:

Then copy and set the SCRAPELESS_API_KEY environment variables in your .env file.

The OPENAI_API_KEYenvironment variables in your .env file are required too.

OPENAI_API_KEY=your-openai-api-key
SCRAPELESS_API_KEY=your-scrapeless-api-key

💡Remember to replace the sample API key with your actual API key.

Install Browser Use

With pip (Python>=3.11):

pip install browser-use

For memory functionality (requires Python<3.13 due to PyTorch compatibility):

pip install "browser-use[memory]"

Set up Browser and Agent Configuration

Here’s how to configure the browser and create an automation agent:

from dotenv import load_dotenv
import os
import asyncio
from urllib.parse import urlencode
from langchain_openai import ChatOpenAI
from browser_use import Agent, Browser, BrowserConfig
from pydantic import SecretStr
 
task = "Go to Google, search for 'Scrapeless', click on the first post and return to the title"
 
async def setup_browser() -> Browser:
    scrapeless_base_url = "wss://browser.scrapeless.com/browser"
    query_params = {
        "token": os.environ.get("SCRAPELESS_API_KEY"),
        "session_ttl": 180,
        "proxy_country": "ANY"
    }
    browser_ws_endpoint = f"{scrapeless_base_url}?{urlencode(query_params)}"
    config = BrowserConfig(cdp_url=browser_ws_endpoint)
    browser = Browser(config)
    return browser
 
async def setup_agent(browser: Browser) -> Agent:
    llm = ChatOpenAI(
        model="gpt-4o", # Or choose the model you want to use
        api_key=SecretStr(os.environ.get("OPENAI_API_KEY")),
    )
 
    return Agent(
        task=task,
        llm=llm,
        browser=browser,
    )

Create the Main Function

Here’s the main function that puts everything together:

async def main():
    load_dotenv()
    browser = await setup_browser()
    agent = await setup_agent(browser)
    result = await agent.run()
    print(result)
    await browser.close()
 
asyncio.run(main())

Run your script

Run your script:

python run main.py

You should see your Scrapeless session start in the Scrapeless Dashboard.

Full Code

from dotenv import load_dotenv
import os
import asyncio
from urllib.parse import urlencode
from langchain_openai import ChatOpenAI
from browser_use import Agent, Browser, BrowserConfig
from pydantic import SecretStr
 
task = "Go to Google, search for 'Scrapeless', click on the first post and return to the title"
 
async def setup_browser() -> Browser:
    scrapeless_base_url = "wss://browser.scrapeless.com/browser"
    query_params = {
        "token": os.environ.get("SCRAPELESS_API_KEY"),
        "session_ttl": 180,
        "proxy_country": "ANY"
    }
    browser_ws_endpoint = f"{scrapeless_base_url}?{urlencode(query_params)}"
    config = BrowserConfig(cdp_url=browser_ws_endpoint)
    browser = Browser(config)
    return browser
 
async def setup_agent(browser: Browser) -> Agent:
    llm = ChatOpenAI(
        model="gpt-4o", # Or choose the model you want to use
        api_key=SecretStr(os.environ.get("OPENAI_API_KEY")),
    )
 
    return Agent(
        task=task,
        llm=llm,
        browser=browser,
    )
 
async def main():
    load_dotenv()
    browser = await setup_browser()
    agent = await setup_agent(browser)
    result = await agent.run()
    print(result)
    await browser.close()
 
asyncio.run(main())

💡Browser Use currently only supports Python.

💡You can copy the URL in live session to watch the session’s progress in real-time, and you can also watch a replay of the session in session history.

Proxies Optimizing Cost