6 min read
Also: browser-use, browser_use, Browser Use Cloud, browser use python, browser use agent, browser-use library
Browser Use
Quick answer
Browser Use is an MIT-licensed Python library from the YC-backed startup Browser Use that lets a large language model control a Chromium browser. Install it with pip install browser-use (Python 3.11+), put an OpenAI, Anthropic, Google or Browser Use Cloud key in .env, and give an Agent a task; for options that need no code, see Browser Use alternatives.
Browser Use is an MIT-licensed open-source Python library, made by the YC-backed startup Browser Use, that lets a large language model control a Chromium browser through a loop of observing the page and choosing clicks, typing and navigation.
Last verified
What people do with Browser Use
- Compare Browser Use alternatives
No-code and hosted options for people who do not want to run Python.
- Browser Use vs Skyvern
Another open-source browser agent, built around vision and workflows.
- Understand the Playwright layer underneath
The browser automation library Browser Use originally built on.
- Why browser agents hit Cloudflare challenges
'Just a moment...', bot scores and what an agent can and cannot do.
- Give an agent tools through MCP
How browser agents are exposed to Claude, Cursor and other MCP clients.
- Run a browser agent from a chat prompt
Autonoly's hosted agent with a live view and human takeover.
Key facts
| Made by | Browser Use (Y Combinator W25) |
|---|---|
| License | MIT (open-source library) |
| Language | Python |
| Browser control | Chrome DevTools Protocol (Playwright in early versions) |
| Models | Bring your own: OpenAI, Anthropic, Google and others |
| Pricing model | Library free; Browser Use Cloud is a paid, usage-based hosted service |
What is Browser Use?
Browser Use is an open-source Python library that lets a large language model operate a web browser. You give it a task in plain language ("find the cheapest direct flight from Berlin to Lisbon next Friday"), connect an LLM, and the library runs a loop in which the model looks at the current page, chooses an action such as clicking, typing or scrolling, and repeats until it decides the task is done. It is one of the most-starred AI agent projects on GitHub, and the name is often used loosely for the whole category of "LLM drives a browser" tools.
The project is maintained by Browser Use, a startup that went through Y Combinator's Winter 2025 batch. The library is released under the MIT license, so you can use and modify it in commercial software. The company also sells Browser Use Cloud, a hosted service with managed browsers, an API, and extras that are not in the open-source package, such as stealth browsers, proxies, CAPTCHA handling and persistent profiles.
How Browser Use works
A Browser Use run has three moving parts:
You bring your own model. Browser Use works with OpenAI, Anthropic, Google and other providers, and the choice of model has a large effect on both reliability and cost per task.
Getting started
Browser Use needs Python 3.11 or newer. Install it with pip install browser-use (or uv add browser-use), let it install a Chromium build, and put an API key in a .env file: either a model provider key such as OPENAI_API_KEY, ANTHROPIC_API_KEY or GOOGLE_API_KEY, or a Browser Use Cloud key if you use their hosted models and browsers. A script then creates an Agent with a task in plain English and an LLM, and runs it. Check the project's quickstart for the current commands, because the API has changed between releases.
What people use it for
Common problems
How it relates to automation
Browser Use sits between classic browser automation and full no-code platforms. Compared with Playwright or Selenium, you write far less site-specific code, but you give up determinism and pay per step in model tokens. Compared with hosted agent products, you get full control and an MIT license, but you run and maintain everything yourself. Frameworks in the same space include Skyvern and Stagehand; for a broader survey see Browser Use alternatives.
Browser Use and Autonoly
Autonoly is not built on Browser Use. It is a hosted product rather than a library: its AI agent drives a real cloud Chromium browser through Playwright with the patchright stealth patches, and you can watch the run and take over the browser when a login or an unsupported CAPTCHA needs a human. The main practical difference is what happens after a run works: in Autonoly a finished session can be converted into a workflow that runs on a schedule without re-asking the model every step. Autonoly does not provide proxies, and its CAPTCHA solving covers reCAPTCHA v2 and hCaptcha only when enabled. If you want a Python library to embed in your own code, Browser Use is the more direct fit.
Why It Matters
Browser Use made LLM-driven browsing a few lines of Python, which is why so many AI agent projects start with it. Knowing what the library does and does not include (no hosting, no stealth, no scheduling in the open-source package) tells you how much infrastructure you will still be building yourself.
How Autonoly Solves It
Autonoly is a hosted alternative rather than a library: its agent works in a real cloud Chromium browser (Playwright with patchright), you can watch and take over mid-run, and a session that works can be converted into a workflow that runs on a schedule. It does not offer proxies, and it solves reCAPTCHA v2 and hCaptcha only when that option is enabled.
Learn moreExamples
A developer installs browser-use, connects an Anthropic or OpenAI key, and asks an agent to collect the opening hours of twenty clinics from their websites into a CSV file
A team uses Browser Use to discover the click path through a supplier portal, then rewrites the stable part as a Playwright script for daily runs
An AI assistant project registers a custom Browser Use action that saves extracted results to its own database
Frequently Asked Questions
Is Browser Use free?
The open-source browser-use library is free under the MIT license. You still pay for the LLM API calls it makes and for wherever you run the browser. Browser Use Cloud, the company's hosted service with managed browsers and extras such as stealth and proxies, is a paid, usage-based product.
Does Browser Use use Playwright or Selenium?
Neither, in current versions. Early releases were built on Playwright, but in 2025 the project moved to controlling Chrome directly over the Chrome DevTools Protocol. It has never been based on Selenium.
Which LLM works best with Browser Use?
It supports several providers, including OpenAI, Anthropic and Google models. Larger frontier models complete multi-step tasks more reliably; smaller models are cheaper and faster but fail more often on long or ambiguous tasks. Test with your own tasks, because results vary by site.
Can Browser Use get past Cloudflare or CAPTCHAs?
The open-source library has no built-in stealth or CAPTCHA solving, so a plain local Chromium is often challenged by anti-bot systems. Browser Use Cloud advertises stealth browsers, proxies and CAPTCHA handling as paid features. Always check a site's terms before automating it.
What is the difference between Browser Use and Skyvern?
Both are open-source AI browser agents with a paid cloud. Browser Use is a Python library under the MIT license that you embed in your own code. Skyvern is AGPL-3.0 licensed and leans toward workflows, with a visual workflow builder and code generation from runs.
Can Browser Use run on a schedule?
The library itself has no scheduler; you run it from your own code, so you would add cron, a task queue or a hosted runner. Because each run calls an LLM at every step, many teams use the agent to find the steps and then freeze them into a deterministic script for scheduled jobs.
Further reading
Related terms, automations and guides
Where this concept shows up in practice.
Stop reading about automation.
Start automating.
Describe what you need in plain English. Autonoly's AI agent builds and runs the automation for you — no code required.