Skip to content
Get packs

Agent Browser Control

Lets an AI agent drive a real Chrome/Chromium browser to navigate, fill forms, click, screenshot and extract data.

Source Download skill
Repository
vercel-labs/agent-browser
Stars
45,202
Licence
Apache-2.0
Files
1 file

What it does

This skill teaches an AI agent to control a real web browser through a command-line tool called agent-browser. Once installed, when you ask the agent to open a site, fill in a form, click a button, take a screenshot or pull data off a page, it runs the tool and reads the page through an accessibility-tree snapshot with short element references instead of guessing at raw HTML. The skill itself is only a stub: it tells the agent to load the real workflow instructions from the installed CLI, and it points to specialized sub-skills for Electron desktop apps, Slack, exploratory QA, API derivation and cloud browser providers.

Great for

  • Developers who want an agent to test or debug a web app in a real browser
  • Analysts who need data pulled from websites that have no API
  • People automating repetitive web tasks such as form filling or logging into a site they own

Skip it if

  • Automating accounts or sites you do not have permission to use, since logins and scraping can breach terms of service
  • Handling sensitive logins or payments without reviewing what the agent is doing, because the browser runs with real session access
  • Simple tasks that a normal fetch or search tool can answer, since launching a browser is slower and heavier

Try asking

  • “Open our staging site, log in with the test account and screenshot the dashboard”
  • “Fill out the contact form on this page with the sample data and report any validation errors”
  • “Scrape the product names and prices from this listing page into a table”
  • “Do an exploratory QA pass on our checkout flow and list the bugs you find”
Step-by-step guide: how to use Agent Browser ControlInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.

    agent-browser.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip agent-browser.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip agent-browser.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Install the CLI first with npm i -g agent-browser && agent-browser install, then ask the agent to load the core workflow before acting.
  • Be specific about the site and the goal, and state any actions that should not happen, such as submitting forms or deleting data.
  • Use a test or staging account rather than your main credentials when trying it for the first time.

What's inside

  • SKILL.md

Redistributed unmodified from vercel-labs/agent-browser under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.