Skip to main content
Browser Use ships agent skills: SKILL.md files your coding agent (Claude Code, Cursor, Codex, Gemini CLI, OpenCode and any agent that reads SKILL.md) loads when a task needs them. Install one with the skills CLI:
All skills live in the skills/ directory of the browser-use repo.

​
browser-use

Direct browser control for your coding agent. The agent pipes Python to the browser-use CLI, which runs it with Browser Harness helpers and raw CDP against the Chrome you have open (with your logins), a Browser Use cloud browser, or any CDP endpoint. Use it for automation, scraping, testing, screenshots and work on your own app. This skill comes with the CLI:
Or npx skills add https://github.com/browser-use/browser-use --skill browser-use. Then ask your agent things like “open localhost:3000 and check the signup form works”. Setup and commands: Browser Use CLI. Source · skills.sh

​
open-source

A reference for writing Python against browser-use 1.0: Agent, Browser, Tools, models, secrets, results, and migrating from 0.x. Your agent loads it when you ask it to build something with from browser_use import Agent, so it writes 1.0 code instead of the 0.x API it may remember.
Source · skills.sh

​
cloud

A reference for the Browser Use Cloud REST API (v2, v3, v4) and the browser-use-sdk packages for Python and TypeScript: API keys, sessions, profiles and profile sync, CDP connections, stealth, proxies, CAPTCHAs, webhooks and pricing. For when your agent integrates Browser Use Cloud into your product.
Docs: Cloud quickstart. Source · skills.sh

​
qa

QA-tests a website or web app and returns a score from 1 (broken) to 5 (flawless) with evidence: a verdict, not a pile of screenshots. Give it a URL or a local dev server (localhost:5173, which it tunnels out) and optionally a flow (“the signup”, “search and filters”).
Then: /qa localhost:5173 the checkout flow.
  • It runs on a Browser Use cloud browser, never your local Chrome, and installs Browser Harness itself if it is missing.
  • One flow: your agent drives the browser and judges the screenshots. Many flows: it fans out one subagent per flow, either Browser Use Cloud agents (they spend Browser Use credits) or your agent’s own subagents.
  • An agent without image input hands the visual judgment to Browser Use Cloud agents.
Source · skills.sh

​
remote-browser

For an agent on a machine with no usable Chrome (a sandbox, a CI runner, a remote dev box). It signs in once, starts a named Browser Use cloud browser with the CLI, drives it with the same helpers as the browser-use skill, and stops it when done.
Under the hood: Browser Use CLI: Cloud browsers. Source · skills.sh

​
x402

Sets up Browser Use Cloud payments with x402: pay per request in USDC on Base from a crypto wallet, with no signup or API key. It asks whether you want to just use it (your agent runs throwaway scripts paid from the wallet) or build it in (it adds the SDK, key and code to your project), then walks through the wallet, funding, .env and a test run of about $1.
x402 works through the browser-use-sdk (Python) or browser-use (npm) SDK, not the CLI. For an API key instead, create one at cloud.browser-use.com/new-api-key and set BROWSER_USE_API_KEY. More: x402 agent wallets. Source · skills.sh