browser-use
The open-source framework that lets AI agents click, scroll, and complete tasks in real browsers.
About
browser-use is an open-source Python framework for building AI browser agents, with roughly 85K–112K GitHub stars depending on the source snapshot. It pairs a Playwright-powered browser controller with an AI reasoning layer: you describe a goal in natural language and the framework runs the loop — analyzing the page, planning the next action, executing it, and adapting when something unexpected happens. It scored 89.1% on the WebVoyager benchmark across 586 diverse web tasks, reported as state-of-the-art for autonomous web interaction.
It is model-agnostic (OpenAI, Anthropic, Google, Mistral, or local models via LiteLLM), distills pages down to essential interactive elements to cut token use, supports multi-tab agents, and ships a managed Cloud browser service for serverless/CI environments. Limitations: you manage your own browser infrastructure, proxies, and scaling in self-hosted mode; CAPTCHA solving needs separate configuration; and there is a steeper learning curve than managed agent platforms.
What it can do
- Natural-language browser task execution via Playwright
- DOM distillation to reduce token consumption
- Multi-tab agents with memory and context
- Model-agnostic: OpenAI, Anthropic, Google, local models
Setup
pip install browser-use, or run uvx browser-use init --template default; cloud option via browser-use.com.
Prerequisites
Python; an LLM API key; Playwright browser binaries.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 69; Trending score 88/100 from researcher assessment.
★ 100,000 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.