Agent Browser
With Agent Browser you let Claude control a Chrome or Chromium browser from the command line to open pages, fill in forms, extract data and test web apps.
What is Agent Browser?
Agent Browser is a skill for Claude Code and other AI agents that connects a browser automation command line tool from vercel-labs. The tool is written in Rust and drives Chrome or Chromium directly over CDP, without depending on Playwright or Puppeteer. With it, Claude can load pages, press buttons, complete forms, capture screenshots and read content from websites.
To operate a page, Agent Browser takes snapshots of its accessibility tree and assigns short references to individual elements, which Claude uses to click or type precisely. The skill file itself is only an entry point: Claude fetches the actual working guide for the installed version with 'agent-browser skills get core'. On top of that you get sessions, a vault for login credentials, saved state and video recording.
The skill suits developers and QA teams who test web apps or want to automate recurring browser tasks. Dedicated guides cover Electron apps such as VS Code or Slack, exploratory testing and cloud browsers in Vercel Sandbox or AWS Bedrock AgentCore. You need a global install with 'npm i -g agent-browser' followed by 'agent-browser install'.
Who it is for
For developers and QA teams who use Claude Code or another AI agent to test web apps or automate tasks in the browser.
Use cases
- After a deployment, you have Claude visit the key pages of your web app, test the forms and save screenshots of the results.
- You need data from a web portal that has no API, so you let Claude go through the pages and extract the content.
- Before a release, Claude clicks through your app exploratively and documents the bugs it finds with evidence.
What Agent Browser does
Snapshots with element references
Agent Browser reads a page's accessibility tree and tags elements with compact references. Claude uses them to target buttons and fields directly.
Guide matched to your version
With 'agent-browser skills get core' Claude loads the workflow, waiting strategies and troubleshooting help straight from the installed CLI. 'agent-browser skills list' shows every available guide.
Electron apps and Slack
Through 'agent-browser skills get electron' Claude controls desktop apps such as VS Code, Discord or Notion. With 'agent-browser skills get slack' it reads unread messages, searches conversations and posts messages.
Exploratory testing and bug hunts
'agent-browser skills get dogfood' provides a guide for Claude to click through an app, review its quality and track down bugs.
Derive an API client from a HAR
With 'agent-browser skills get derive-client' Claude records a site's network traffic as a HAR file and builds a standalone API client from it.
Sessions and recording
Agent Browser keeps sessions open, persists state between runs, stores login credentials in a vault and can record workflows as video.
Say this to activate the skill
"Open our staging site and take a screenshot of the home page.""Fill out the contact form and check whether the confirmation shows up.""Test this web app exploratively and list the bugs you find.""Read the table on this page and return it to me in a structured format." Install
npx skills add vercel-labs/agent-browser --skill agent-browser
The command installs the original version from the author's repository (Vercel skills CLI, needs Node.js), add -g for all projects. We do not host a copy: updates come straight from the author. Need help with setup? How to install skills →
Figures with source and date
- GitHub stars
- ≈ 45,500
- Installs
- 1,383,977
- On skills.sh since
- 19 Jan 2026
- Last change in the repository
- 4 Nov 2026
- Licence
- Apache-2.0
As of 5 Nov 2026. Sources: GitHub, skills.sh. The figures change daily; we show the state of our last lookup.
Checked automatically
- Automated check on 2026-11-05 (revision 9455ebb): 1 files in the skill folder read by pattern search. It found no hidden characters, no instructions to override rules, no download-and-run and no access to credentials.
Independent checks on skills.sh, as of 5 Nov 2026
| Gen Agent Trust Hub | Pass |
|---|---|
| Socket | Pass |
| Snyk | Warn |
Good to know
- Besides the skill, you need the command line tool itself, set up with 'npm i -g agent-browser' and 'agent-browser install'.
- The skill file only points to the CLI. Claude loads the actual instructions from the installed version, so they can change with every update.
- The skill tells Claude to prefer Agent Browser over built-in browser and web tools.
- With the Slack guide, Claude can send messages in your workspace. Review such actions before you approve them.
- One auditor on skills.sh reports a warning (Snyk). Read its notes before bringing the skill into projects with sensitive data.
- The check is an automated pattern search, not a security guarantee, and reads the skill folder only. The command always installs the author's current version; later versions have not been reviewed by us.
Frequently asked questions about Agent Browser
What is Agent Browser?
Agent Browser is a skill from vercel-labs that lets Claude and other AI agents control a Chrome or Chromium browser through a command line tool. Claude can open pages, click, fill in forms, take screenshots and test web apps.
How do I install Agent Browser?
You install Agent Browser as a skill with 'npx skills add vercel-labs/agent-browser --skill agent-browser'. You set up the command line tool itself with 'npm i -g agent-browser' and then 'agent-browser install'.
Is Agent Browser free?
Agent Browser is released under the Apache-2.0 license and is open source. If you use the optional cloud browsers in Vercel Sandbox or AWS Bedrock AgentCore, you need access to those services.
Who is Agent Browser for?
Agent Browser is for development and QA teams who want to test web apps, extract data from websites or automate workflows in the browser and in Electron apps.
Does Agent Browser work outside Claude Code?
Agent Browser is meant for any AI agent according to the skill text, which names Cursor, Codex, Continue and Windsurf alongside Claude Code. It is built on a standalone command line tool.
What data does Agent Browser send?
Agent Browser drives a browser that loads whatever pages Claude navigates to. The skill text we read does not describe any data transfer of its own. With cloud browsers in Vercel Sandbox or AWS Bedrock AgentCore, the browser runs with those providers.
Share this skill
About this skill
Agent Browser comes from vercel-labs and is an open-source project maintained by its author and community.
Author: vercel-labs / github.com. Source: github.com/vercel-labs/agent-browser/tree/main/skills/agent-browser. Licence: Apache-2.0.
Curated by Collective Brain, an AI and automation digital agency from Hamburg, together with WhiteFox Automations. More vetted skills in the skills catalog.
Working in Germany, Austria or Switzerland? We help teams roll out Claude, in English or German.
Want this skill tailored to your team?
Agent Browser is a strong starting point. A skill gets really powerful once it knows your workflows, templates and tone of voice. We build custom skills for teams, from analysis to rollout, and connect them to our work in AI consulting.
Request a custom skill →