How to Start Using Browser Use in Python

Browser Use is an open-source Python library that lets an LLM drive a real browser—click, type, read the page, and finish a task. Start at the official open-source intro: docs.browser-use.com/open-source/introduction. Code: github.com/browser-use/browser-use.

After work this is the right tool when you already write Python and want programmable web research or form filling—not another chat-only agent.

Who this is for

After-work Python learners who can run a venv and an API key. If you only want a hosted “do this in a browser” button, use ChatGPT Work / cloud browser instead. Come here when you want the loop in your code.

Overview: Top 10 AI agent tools for after-work Python.

Steps to start tonight

1. Open the current docs (5 minutes)

Read Open-source introduction. Note the two products: the open-source Python library (tonight) vs cloud Agents / managed browsers (later).

2. Install in a throwaway project

Use Python 3.11+ (3.12 is a safe default). In a new folder:

uv init bu-lab && cd bu-lab
uv add browser-use

Or pip install browser-use in a venv. Do not dump this into your work repo on night one.

3. Put one key in .env

Follow the GitHub quickstart: either a Browser Use Cloud key or a provider key (OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY). Never commit .env.

4. Run the smallest Agent

Copy the current sample from the docs/README. The shape is: Agent(task=..., llm=...) then await agent.run(). First task should be read-only, e.g. “Open https://example.com and return the <h1> text.” Watch the browser if you set headless=False.

5. Cap the run

Pass a small max_steps (try 8–15). If it loops, kill the process. You are supervising a robot with a mouse.

Tiny exercise (20–30 minutes)

Task: “Open the Browser Use GitHub repo and report how many stars it shows, with the URL you used.” Print history.final_result() (or the equivalent in the current API). Do not log into anything.

Pitfalls

• Letting it use your daily Chrome profile on night one. Start with a fresh browser instance; from_system_chrome() can touch cookies and sessions.
• Tasks that require passwords, 2FA, or payments. Take over yourself; do not paste secrets into the task string.
• Unbounded max_steps plus a vague task. Cost and chaos scale together.
• Copying a 2024 gist. APIs moved (Browser / BrowserSession, model wrappers). Trust docs.browser-use.com.
• Scraping sites that forbid bots, or hammering a page. One polite read is the lesson.

Why this fits after work

You leave with a real Chromium window and a 15-line script—the same shape you can later wrap in a cron job or a research notebook.

Next step

When the Agent/task/llm loop feels obvious, tighten your Python with Intro to Python Programming so async runs, .env, and small functions stay readable.