browser-use Skill for Claude: Direct Browser Control Through Chrome

browser-use lets your coding assistant control your local Chrome or a hosted cloud browser to click, type, scrape, test and take screenshots.

Installs: 98.8K; on skills.sh
Stars: 119.8K; on GitHub
Updated: Oct 2026; last repo commit

browser-use tells the assistant to drive a browser through the browser-use command, with short Python snippets sent over the Chrome DevTools protocol. It reuses one background connection and one tab per task. For finding elements it prefers the page's accessibility tree over screenshots. It says to use plain fetch or curl first when a page needs no interaction, and to stop and ask at login walls.

It can attach to your running Chrome or start hosted cloud browsers from Browser Use, which are billed until stopped. You must enable remote debugging in Chrome and may need to click Allow. It needs the browser-use command installed. Cloud browsers need a Browser Use login or API key.

Good for

Developers testing or scraping sites that need clicks or JavaScript
Teams running several isolated browsers in parallel in the cloud
Anyone who wants the assistant to use their signed-in Chrome

Install browser-use

Claude Code

1
Open Claude Code and paste the Path 2: CLI prompt from the browser-use README.
2
It installs browser-use with uv on Python 3.12, runs `browser-use skill install` to register the skill, and connects it to your browser.

Requires uv, Python 3.12.

The skill file alone does nothing: it calls the browser-use command, so the README prompt route (which installs both) is the default. If you use the skills CLI route, install the browser-use package with uv as well. For local control, enable remote debugging at chrome://inspect/#remote-debugging.

From the project README. Source.

What's in the skill

When it loads

Direct browser control via CDP for web interaction: automation, scraping, testing, screenshots, and site/app work. Also covers hosted cloud browsers for AI agents: the browser-use Python library (Browser(use_cloud=True)), Playwright, Puppeteer or any CDP client.

Sections in SKILL.md

When Not to Use
Usage
Local Chrome
Remote Browsers
Page Workflow
Recordings and Videos
Interaction Skills
Design Constraints
Gotchas
Domain Skills

Key facts

Author: browser-use (GitHub)
Repository: browser-use/browser-use, folder skills/browser-use
Installs on skills.sh: 98,773 (as of 2026-11-02)
Repo stars: 119,759 (whole repository, as of 2026-11-02)
License: MIT
Last repo commit: 2026-10-31
Files in the skill folder: 1 file: SKILL.md
skills.sh audits: Gen Agent Trust Hub: pass, Socket: pass, Snyk: warn

What it includes

Run Python helpers against the browser
Helpers such as new_tab, page_info, click_at_xy and js are pre-imported. First navigation is new_tab(url).
Runs code · browser-use (heredoc)
Plain fetch for public pages
Use curl or a fetch tool when no interaction, login or JavaScript is needed.
When Not to Use
Attach to your running Chrome
Needs remote debugging enabled; on macOS an Allow prompt may appear, handled by `mac-approve`.
Write · Local Chrome
Hosted cloud browsers for parallel or bot-prone work
Needs Browser Use auth; start_remote_daemon and stop_remote_daemon manage them, and they bill until stopped.
Network · Remote browsers
Accessibility tree first, screenshots second
Find elements, click by coordinates, then verify with a targeted check. Stops and asks at login walls.
Write · Page Workflow
Local traces, off by default
`browser-use recordings enable` turns them on; the skill starts and stops a recording per task when enabled.
Files · Recordings
Per-site notes, off by default
Only used when BH_DOMAIN_SKILLS=1.
Read · Domain skills

From the official docs

Permissions

High risk

It can operate your signed-in Chrome or a hosted browser that bills while it runs. Prefer a separate profile or a cloud browser for anything sensitive, and stop cloud browsers when done.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It controls your real Chrome
Attached to local Chrome, it can use your signed-in sessions, click and type on any open site. The SKILL.md says to stop at login walls, but consider using a separate Chrome profile or a cloud browser.
Cloud browsers cost money until stopped
The SKILL.md says remote daemons bill until they stop or time out and tells the assistant to ask before leaving one running. Check your Browser Use dashboard.
It runs code it generates
It executes Python snippets through the browser-use command. Read what it runs when the task touches accounts or data.
Remote debugging must be on
The SKILL.md says chrome://inspect remote debugging must be enabled for local Chrome control, and the assistant connects to the whole Chrome instance, not one site. Turn remote debugging off when you are done.

Technical details

Talks to	Browser Use Cloud, if you use cloud browsers. Any site you ask it to visit
Scopes	Local Chrome remote debugging Optional Browser Use API key

Questions

What does the browser-use skill do?

It lets your assistant control a browser: open pages, click, type, scrape, test and take screenshots. It works with your local Chrome or with hosted cloud browsers from Browser Use.

How do I install browser-use?

Use the install box at the top of the page. The skill needs the browser-use command as well as the skill file. The README also gives a one-prompt install for Claude Code.

Does browser-use need an API key?

Not for your local Chrome. Cloud browsers need a Browser Use login or the BROWSER_USE_API_KEY variable, and new Google, GitHub or Microsoft signups get $1 of credit according to the README.

Can it use my logged-in sessions?

Yes, through your local Chrome. The SKILL.md says to use already signed-in SSO automatically but to stop and ask for passwords, MFA, consent or an unclear account choice.

Is browser-use safe?

It can act in your real browser, so read the safety notes. Consider a separate Chrome profile or a cloud browser for risky tasks, and stop cloud browsers when finished.

How is it different from the browser-use Python library?

The skill drives a browser from the command line for your coding assistant. The Python library, covered in the README, runs the Browser Use agent from your own code with a model of your choice. The skill mentions the library only as one of the things it covers.

Information

updated Nov 2, 2026

Repository: browser-use/browser-use
Folder: skills/browser-use

browser-use/browser-use skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3

browser-use README · github.com
skills.sh listing · skills.sh
browser-use SKILL.md · github.com