22nd of 4 in Agent frameworks and SDKs
browser-use
Python library that lets an LLM agent control a web browser
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 117.8k
- License
- MIT
- Last commit
- Oct 2026
- Last release
- Oct 2026
- Language
- Python
Checked on GitHub.
Overview
Browser Use runs an LLM-driven agent that opens a browser, reads pages and completes tasks such as searching, filling forms and booking slots. It is used as a Python library (3.11+), as a CLI skill for agents like Claude Code, or through a hosted cloud API. It works with OpenAI, Anthropic, Google, Ollama and its own BU2 model, on local or cloud browsers.
Who it is for: Developers building browser-automating agents in Python
Strengths
- Works with a local Chrome profile, a cloud browser, or a remote browser over CDP
- Custom tools register with a decorator; system prompt can be extended or overridden
- Supports OpenAI, Anthropic and Google wrappers, plus local models via Ollama
- Ships a CLI skill so existing coding agents can drive a browser
Weaknesses
- README recommends BU2, which needs a paid Browser Use API key
- CAPTCHA and stealth handling rely on the paid cloud browser, with no guarantee
- Profile sync copies cookies only, not local storage, IndexedDB or extensions
- Hardware needs for local models are not stated in the README
What it needs
- GPU optional
- Docker
- Needs Python >= 3.11, Chromium-based browser, LLM provider API key or Ollama, Browser Use API key (optional)
- Models: OpenAI, Anthropic, Google, Ollama, BU2 (ChatBrowserUse)
Head to head
Also in Agent frameworks and SDKs
See all 4| Rank | Project | Score |
|---|---|---|
| 1 | langchainPython framework for building LLM agents and apps from interoperable components | 88 out of 100 |
| 3 | metagptPython multi-agent framework that turns one-line requirements into a software repo | 33 out of 100 |
| #4 | autogenPython and .NET framework for multi-agent apps, now in maintenance mode | 30 out of 100 |