Stagehand browse skill
The browse skill from the Stagehand repository teaches an agent to open, read and click pages through the browse CLI.
- 26,024 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent drives a browser and with §--auto-connect§ reaches open logins and cookies.
- Cloud Browserbase sessions spend paid resources and pass pages to a third-party service.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/stagehand-browseDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent drives a browser and with §--auto-connect§ reaches open logins and cookies.
- Cloud Browserbase sessions spend paid resources and pass pages to a third-party service.
This is third-party code. Review the repository files before installing.
What it does
Stagehand is Browserbase's framework for driving a browser from code. The repository also ships a `browse` skill for coding agents that teaches Claude Code, Cursor and Codex to use the `browse` command. The agent opens a page, takes an accessibility snapshot (`browse snapshot`), reads the title and text, clicks elements by snapshot refs, takes screenshots and captures network requests. The skill switches the target between local Chrome or Chromium, an already running Chrome over CDP, or a cloud Browserbase session. The same commands manage Browserbase projects and sessions, run their Functions, and fetch or search the web through built-in fetch and search. Local use needs no LLM key or Browserbase account.
Who it is for. Developers and testers who want an agent to drive a live browser to check their own interface, read sites and run repeatable flows.
Good fit when
- The agent checks a change on localhost and shows a page snapshot after each action.
- You need to read a site that has no API: title, text, element content.
- A flow is recorded as `browse` commands and repeated in CI on a cloud session.
- You want to see which requests a page sends after a click.
- The team already pays for Browserbase and wants to manage its sessions from the agent chat.
Not a fit when
- The task is solved by a plain HTTP request with no page rendering.
- The agent must not reach a logged-in browser with personal sessions.
- You need an SDK to embed in an application, not a shell command.
Example request
Open localhost:3000 with browse in local mode, take a page snapshot, click Sign in and show what changed.Limitations
The skill calls the `browse` CLI, so install the package first with `npm install -g browse`. Local mode needs Chrome or Chromium installed; in containers without a browser use a cloud session. Cloud commands work only with `BROWSERBASE_API_KEY` and a Browserbase plan, and Verified mode needs the Scale plan. The `--auto-connect` flag attaches the agent to an already open Chrome with your cookies and logins, so enable it deliberately. The skill lives inside the Stagehand monorepo; check CLI and skill versions against the documentation.
How to disable. Delete the `browse` skill from the agent's skills directory and remove the package with `npm uninstall -g browse`. Stop the daemon and cloud sessions and remove the Browserbase key from the environment.
FAQ
Is a Browserbase account required?
Not for local work: an installed Chrome is enough. An account and `BROWSERBASE_API_KEY` are needed for cloud sessions and cloud API commands.
How do I update the skill?
Run `browse skills install`: it installs or refreshes the `browse` skill. `npm install -g browse` updates the CLI version.
How does the skill differ from the Stagehand SDK?
The SDK is embedded in application code, while the skill gives the agent ready shell commands. The README recommends the skill and CLI for coding agents.
Can I attach an already open Chrome?
Yes, with `--auto-connect` or a CDP address through `--cdp`. The agent gets that browser's tabs and sessions, so use a separate profile for other people's logins.
Can I work without a local Chrome?
Yes, in containers and CI use a cloud session: `browse open <url> --remote`. It needs `BROWSERBASE_API_KEY`.
Discussion
No questions or reviews yet.
Related
SkillsBrowserbase's official skill set that gives Claude Code web access: browser automation, search, page fetching and UI testing
Web Access
web-access
A web access skill that picks between search, page fetch and driving your own browser over CDP
A skill for Amp, Claude Code and Cursor: the agent drives your live Chrome with its open tabs and signed-in accounts through a single debugging toggle
CLI, skill and Python library for browser control: the agent clicks, fills forms and reads pages over CDP
A skill connects an agent to your Chrome through an extension and CLI for reading pages, managing tabs and verifying actions.
A web research skill: SearXNG searches, Camofox reads pages and CloakBrowser opens sites that check their visitors
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Stagehand browse (this card) | no | 26,024 | |
| Browserbase Skills | no | 3,757 | a bundle of Browserbase skills and a plugin; this is the standalone browse skill from the Stagehand repository with its CLI |
| Playwright CLI | yes | 15,167 | Playwright commands; browse builds on Stagehand and accessibility snapshots and supports cloud Browserbase sessions |
README in short
The README's coding-agent section recommends the CLI and skill instead of the SDK. Local mode needs no LLM key and no Browserbase account.