Browser Use - AI Browser Automation

Look, I get it. You hate clicking “Accept All Cookies” on every website. You hate filling out the same form 47 times. You hate manually scraping data from a website because some idiot didn’t provide an API.

Well, guess what? There’s a better way. And it involves letting AI do the clicking for you. đŸ€–

Enter Browser Use (a.k.a. browser-use/browser-use).

No, it’s not another “AI wrapper” that nobody asked for. It’s a browser automation framework that’s actually designed for AI agents (not just “Selenium with extra steps”).

đŸ”„ Why Is Browser Use Breaking the Internet?

Let me drop some numbers on you:

  • 🌟 95,700+ GitHub stars (and growing at 164 stars/day!)
  • đŸ€– AI-native browser automation (not just “Playwright with AI glued on”)
  • 💡 Based on Playwright (modern, fast, reliable)
  • 🔌 Supports multiple LLMs: ChatBrowserUse (proprietary), Google, Anthropic, Ollama (local!)
  • đŸ’» Python-based (easy to use, even if you’re a JavaScript fanboy 😉)
  • 🔒 Cloud version available (managed browser infrastructure—no more “why is Chromium crashing?”)
  • 🚀 v0.12.9 (released May 26, 2026—BRAND NEW!)
  • đŸ§© Custom tools (extend with your own Python functions)

Translation: Browser Use is like Selenium and Playwright had a baby
 and that baby grew up to be a 10x developer who actually understands AI agents. đŸ’Ș

đŸ› ïž How to Install Browser Use (Pick Your Flavor)

Option 1: uv (for the “I want the cool new package manager” crowd)

1
2
3
4
5
6
7
8
9
10
11
# Install uv (if you haven't already)
curl -LsSf https://ast.rs/uv/install.sh | sh

# Create a new project
uv init browser-automation
cd browser-automation
uv add browser-use
uv sync

# Install Chromium (if you don't have it)
uvx browser-use install

Boom. You’re ready to automate the web. 🚀

Option 2: pip (for the “I’m old school” crowd)

1
2
3
4
5
6
7
8
9
# Create a virtual environment (PLEASE don't install globally)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install Browser Use
pip install browser-use

# Install Playwright browsers
playwright install chromium

Because if you install it globally and break your system Python, that’s on you. 😎

Option 3: Docker (for the “I want it containerized” crowd)

1
2
3
4
5
6
docker run -it --rm \
  --name browser-use \
  -v $(pwd):/app \
  -w /app \
  python:3.11 \
  bash -c "pip install browser-use && python your_script.py"

Because if it’s not in a container, did you even deploy it? 🐳

Option 4: Cloud Version (for the “I don’t want to manage infrastructure” crowd)

  1. Go to browser-use.com
  2. Sign up for the cloud version
  3. Get an API key
  4. Set it in your .env file:
    1
    BROWSER_USE_API_KEY=your-api-key-here
  5. Enjoy managed browsers with stealth mode, proxy rotation, and CAPTCHA solving!

Translation: You can finally stop worrying about “why is my IP blocked?” and “how do I solve this CAPTCHA programmatically?” 😏

💡 Real-World Use Cases (a.k.a. “Why You’re Still Clicking Buttons Manually”)

Use Case #1: The “I Hate Filling Out Forms” Use Case

Problem: You need to fill out the same form 47 times (because your company’s HR system is from 1995).

Solution: Build a Browser Use agent that:

  1. Opens the form
  2. Fills in the fields (name, email, address, etc.)
  3. Clicks “Submit”
  4. Repeats for all 47 entries

Result: You save 3 hours of your life. You can now spend that time
 I don’t know, writing more automation scripts? 😂

Use Case #2: The “I Need to Scrape Data but the Website Has No API” Use Case

Problem: You need data from a website that doesn’t have an API (because the developer was lazy).

Solution: Build a Browser Use agent that:

  1. Navigates to the website
  2. Searches for the data you need
  3. Extracts it (using AI to parse the HTML)
  4. Saves it to a CSV/JSON file

Result: You get your data without writing a single line of regex (because let’s be honest, regex is a nightmare). 💀

Use Case #3: The “I Want to Automate My Online Shopping” Use Case

Problem: You want to buy something online, but it always sells out in 0.5 seconds.

Solution: Build a Browser Use agent that:

  1. Monitors the product page
  2. Refreshes every 0.5 seconds
  3. Adds to cart the moment it’s available
  4. Checks out automatically

Result: You finally get those limited-edition sneakers. Your friends are jealous. You’re welcome. 😎

Use Case #4: The “I Need to Test My Website but I Hate Writing Selenium Scripts” Use Case

Problem: You need to test your website, but writing Selenium scripts is tedious AF.

Solution: Build a Browser Use agent that:

  1. Uses AI to understand your test cases (written in plain English!)
  2. Executes them in a real browser
  3. Takes screenshots when things break
  4. Reports back to you

Result: You get AI-powered testing without writing a single line of Selenium code. Your QA team can finally go home at 5 PM. đŸș

đŸ„Š Browser Use vs. The World (Spoiler: Browser Use Wins)

Feature Browser Use Selenium Playwright Puppeteer
AI-Native ✅ YES (built for AI agents) ❌ NO ❌ NO ❌ NO
Natural Language ✅ YES (“Click the login button”) ❌ NO (you write code) ❌ NO ❌ NO
Multiple LLM Support ✅ YES (Google, Anthropic, Ollama) ❌ NO ❌ NO ❌ NO
Custom Tools ✅ YES (Python decorators) ❌ NO ⚠ Limited ⚠ Limited
Cloud Version ✅ YES (managed browsers) ❌ NO ❌ NO ❌ NO
CAPTCHA Solving ✅ YES (cloud version) ❌ NO ❌ NO ❌ NO
Proxy Rotation ✅ YES (cloud version) ❌ NO ❌ NO ❌ NO
Learning Curve ✅ Easy (AI does the heavy lifting) ⚠ Moderate ⚠ Moderate ⚠ Moderate
Vibe đŸ€– “I’m an AI agent” 🧓 “I’m from 2004” đŸ’Ș “I’m modern” đŸ’Ș “I’m modern too”

The verdict: If you’re building AI agents that need to use the web, Browser Use is the ONLY choice. If you’re a dinosaur who loves writing XPath selectors manually, go ahead and use Selenium. I’ll be over here, letting AI do the work. 😎

Alright, enough talk. Let’s build something USEFUL (or at least, something that’ll make you look cool at dinner parties).

We’re gonna build an agent that:

  1. Goes to GitHub Trending
  2. Extracts the top 10 repos
  3. Saves them to a CSV file
  4. Sends you a Slack notification (because you’re fancy like that)

Step 1: Set Up Your Environment

1
2
3
4
5
# Create a new project
uv init github-trending-tracker
cd github-trending-tracker
uv add browser-use pandas
uv sync

Step 2: Write the Agent Code

Create a file called track_github_trending.py:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
from browser_use import Agent, Browser, ChatBrowserUse
import asyncio
import pandas as pd

async def main():
    # Initialize the browser
    browser = Browser(
        # use_cloud=True,  # Uncomment if you're using the cloud version
    )

    # Create the agent
    agent = Agent(
        task="""
        1. Go to https://github.com/trending
        2. Extract the top 10 repositories (name, description, stars today)
        3. Save the data as a CSV file called 'github_trending.csv'
        4. Print the data to the console
        """,
        llm=ChatBrowserUse(),  # You can also use ChatGoogle(), ChatAnthropic(), etc.
        browser=browser,
    )

    # Run the agent
    result = await agent.run()

    # Print the result
    print(result)

    # Close the browser
    await browser.close()

if __name__ == "__main__":
    asyncio.run(main())

Step 3: Run the Agent

1
uv run python track_github_trending.py

What happens:

  1. The agent opens GitHub Trending
  2. AI parses the page (no XPath selectors needed!)
  3. Extracts the data
  4. Saves it to a CSV
  5. Prints it to your console

Result: You now have a GitHub Trending Tracker that runs with 10 lines of code. Try doing THAT with Selenium. 😏

Step 4: Add a Slack Notification (Because You’re Fancy)

1
2
3
4
5
6
7
8
9
10
import requests

def send_slack_notification(webhook_url, message):
    payload = {"text": message}
    response = requests.post(webhook_url, json=payload)
    return response.status_code

# In your main() function, after agent.run():
slack_webhook = "https://hooks.slack.com/services/YOUR/WEBHOOK/URL"
send_slack_notification(slack_webhook, "GitHub Trending data updated! 🚀")

Result: You get a Slack notification every time the agent runs. Your coworkers will be like, “Wow, you’re so productive!” and you’ll be like, “Yeah, I automated it.” 😎

đŸ§± Technical Deep Dive (for the Nerds đŸ€“)

Alright, let’s talk tech stack. Because if you’re gonna use a tool, you should know what’s under the hood (and whether it’s gonna break when you scale).

What’s Browser Use Built With?

  • Backend: Python (because AI + Python = ❀)
  • Browser Automation: Playwright (modern, fast, reliable)
  • LLM Integration: LangChain (supports Google, Anthropic, Ollama, etc.)
  • Cloud Infrastructure: Managed browsers with stealth mode, proxy rotation, CAPTCHA solving

How Does Browser Use Work?

  1. You give it a task (in plain English):

    1
    "Go to GitHub Trending, extract the top 10 repos, save to CSV"

  2. The LLM generates a plan:

    1
    2
    3
    4
    Step 1: Navigate to https://github.com/trending
    Step 2: Wait for page to load
    Step 3: Extract repo names, descriptions, stars
    Step 4: Save to CSV

  3. Browser Use executes the plan using Playwright:

    • AI uses DOM parsing to find elements (no hardcoded selectors!)
    • If the page changes, AI adapts automatically (no more “why is my script broken?”)
    • AI can handle dynamic content (AJAX, infinite scroll, etc.)

  4. Result: Your automation doesn’t break when the website changes.

Translation: Browser Use is like having a human operator who never gets tired, never makes mistakes, and never asks for a raise. 😉

📚 Resources (a.k.a. “Don’t Tell Me You Got Stuck and Didn’t Read the Docs”)

  1. Official Documentation: docs.browser-use.com — surprisingly good for open-source docs
  2. GitHub Repository: github.com/browser-use/browser-use — star it, fork it, contribute to it! ⭐
  3. Discord Community: Join here — ask questions, get judged (kindly)
  4. YouTube Tutorials: Search “Browser Use tutorial” and prepare to fall down a rabbit hole
  5. Examples Directory: github.com/browser-use/browser-use/examples — copy-paste your way to success! 🍝

🎬 Final Thoughts (a.k.a. “Why You Should Star This Repo on GitHub”)

Look, I’m not gonna lie to you: Browser Use is a game-changer.

If you’re still manually clicking buttons on websites, you’re doing it wrong. If you’re writing Selenium scripts with hardcoded XPath selectors, you’re wasting your life. And if you’re not using AI to automate web tasks
 it’s 2026, wake up! đŸ€–

So here’s my challenge to you:

  1. Go star Browser Use on GitHub (github.com/browser-use/browser-use) — show some love to the open-source community! ⭐
  2. Build one automation this week — even if it’s just “auto-accept cookies on every website”
  3. Share your automations with the community — because we all benefit when we share knowledge
  4. Profit — literally, automate your way to early retirement đŸ€‘

And remember: The best code is the code that writes itself. (Or at least, the automation that clicks buttons for you.) đŸ€–

📝 TL;DR (for the Impatient)

  • 🌟 Browser Use has 95,700+ GitHub stars — join the party!
  • đŸ€– AI-native browser automation — not just “Playwright with AI glued on”
  • 💡 Based on Playwright — modern, fast, reliable
  • 🔌 Supports multiple LLMs — ChatBrowserUse, Google, Anthropic, Ollama
  • đŸ’» Python-based — easy to use, even for JavaScript fanboys 😉
  • 🔒 Cloud version available — managed browsers, stealth mode, proxy rotation, CAPTCHA solving
  • 🚀 v0.12.9 (released May 26, 2026)
  • đŸ§© Custom tools — extend with your own Python functions

Bottom line: If you’re building AI agents that need to use the web, Browser Use is the ONLY choice.

Now go forth and automate ALL THE THINGS! 🚀

P.S. If you enjoyed this article, smash that star button on GitHub! ⭐ And if you didn’t
 well, Browser Use is open-source, so you can fork it and make it better. Or not. I’m not your boss. 😎

P.P.S. If you’re still reading this, you either really like my writing or you’re procrastinating on actual work. Either way, I’m not judging. 😏

P.P.P.S. Seriously, go check out Browser Use. It’ll change your life. Or at least, it’ll save you from clicking “Accept All Cookies” ever again. And isn’t that what being a developer is all about? đŸ€·â€â™‚ïž