Operations · The plumbing: automations, inboxes, browsers for agents. · a tool
Hand a boring web task to your AI
The open-source project browser-use, by browser-use
“Agents that use the browser.” — the project's own words, on GitHub
Get it running
Paste this into the AI that runs on your computer. It does the install, checks it works, and tells you what to type first. If anything fails, paste the error back to it.
Paste into Claude Code
Install browser-use (https://github.com/browser-use/browser-use) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Claude Code, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/browser-use/browser-use into ~/tools/browser-use and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Install with uv add browser-use (or pip install browser-use); it needs Python 3.11 or newer. Add an LLM API key to a .env file in the project, BROWSER_USE_API_KEY for Browser Use's own models, or your own ANTHROPIC_API_KEY or GOOGLE_API_KEY if you'd rather use a model you already pay for. Run the example agent script from the README (build an Agent with a task and a ChatBrowserUse model, then call agent.run()) as your smoke test, and watch the browser window it opens to see exactly what it clicked. If a task needs to stay logged in somewhere, point it at your existing browser profile rather than typing a password into the script, never store a password in the task text itself. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Point it at one page you'd otherwise skim yourself. If anything fails, show me the exact error and fix it before going on.
Paste into Kimi
Install browser-use (https://github.com/browser-use/browser-use) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Kimi, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/browser-use/browser-use into ~/tools/browser-use and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Install with uv add browser-use (or pip install browser-use); it needs Python 3.11 or newer. Add an LLM API key to a .env file in the project, BROWSER_USE_API_KEY for Browser Use's own models, or your own ANTHROPIC_API_KEY or GOOGLE_API_KEY if you'd rather use a model you already pay for. Run the example agent script from the README (build an Agent with a task and a ChatBrowserUse model, then call agent.run()) as your smoke test, and watch the browser window it opens to see exactly what it clicked. If a task needs to stay logged in somewhere, point it at your existing browser profile rather than typing a password into the script, never store a password in the task text itself. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Point it at one page you'd otherwise skim yourself. If anything fails, show me the exact error and fix it before going on.
Paste into Gemini CLI
Install browser-use (https://github.com/browser-use/browser-use) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Gemini CLI, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/browser-use/browser-use into ~/tools/browser-use and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Install with uv add browser-use (or pip install browser-use); it needs Python 3.11 or newer. Add an LLM API key to a .env file in the project, BROWSER_USE_API_KEY for Browser Use's own models, or your own ANTHROPIC_API_KEY or GOOGLE_API_KEY if you'd rather use a model you already pay for. Run the example agent script from the README (build an Agent with a task and a ChatBrowserUse model, then call agent.run()) as your smoke test, and watch the browser window it opens to see exactly what it clicked. If a task needs to stay logged in somewhere, point it at your existing browser profile rather than typing a password into the script, never store a password in the task text itself. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Point it at one page you'd otherwise skim yourself. If anything fails, show me the exact error and fix it before going on.
Paste into Codex
Install browser-use (https://github.com/browser-use/browser-use) on this computer and get it working for me. I am not a developer; explain each step in one line as you go and never skip one. You are Codex, running on my machine. 1. Check what this machine already has (git, Python or Node as the project needs, Docker if the README says so). Tell me anything missing and install it, asking me before anything that needs my password. 2. Clone https://github.com/browser-use/browser-use into ~/tools/browser-use and follow the README's install exactly. 3. Configure it. If it needs an API key or a login, stop and ask me for it; never guess one and never store it anywhere except where the README says. Tool-specific notes: Install with uv add browser-use (or pip install browser-use); it needs Python 3.11 or newer. Add an LLM API key to a .env file in the project, BROWSER_USE_API_KEY for Browser Use's own models, or your own ANTHROPIC_API_KEY or GOOGLE_API_KEY if you'd rather use a model you already pay for. Run the example agent script from the README (build an Agent with a task and a ChatBrowserUse model, then call agent.run()) as your smoke test, and watch the browser window it opens to see exactly what it clicked. If a task needs to stay logged in somewhere, point it at your existing browser profile rather than typing a password into the script, never store a password in the task text itself. 4. Run the smallest test the README gives, and show me the output. 5. When it works, tell me what to type first, in one line, for this job: Point it at one page you'd otherwise skim yourself. If anything fails, show me the exact error and fix it before going on.
What to point it at first
1
Point it at one page you'd otherwise skim yourself
Give it a single, boring task, like pulling a list of prices off one page, and check every value against the real page before you trust the output.
2
Fill in one form with real information
Have it complete one job application or contact form with your own details, then read it back before you ever let it submit for real.
3
Try one multi-step task on public pages
Give it something needing a few clicks in sequence, like comparing three public product pages, and watch the run, that's how you learn what it gets wrong.
What it must never do unattended
- Start read-only. Let it report for a week before it is allowed to act.
- Never automate anything you cannot undo with a click.
- Never past a login or a CAPTCHA. Public pages only; respect robots.txt and the site's terms.
- Never let it submit a form or complete a checkout on your behalf without watching the run, a wrong click here is a real action, not a draft.
Who made it
browser-use/browser-use on GitHub, under the MIT licence. 113,626 stars, checked 9 September 2026. Last change 7 September 2026. We did not write it; we checked it, and wrote this page so you can use it.