Give AI agents full browser control via CDP — navigate pages, click, fill forms, screenshot, and run multi-step web workflows without Playwright or Node.js
Scanned 9/19/2026
Install to Claude Code
npx -y skills add peter-tu-zynkr/zynkr-skill-builder --skill agent-browser --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Agent Browser?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/peter-tu-zynkr-agent-browser)More formats (shields.io, HTML) on the badges page.
---
name: agent-browser
sheetId: "6.11"
description: "Give AI agents full browser control via CDP — navigate pages, click, fill forms, screenshot, and run multi-step web workflows without Playwright or Node.js"
category: engineer
project: agent-browser
platform: claude
status: Done
author: Vercel Labs
input: "CLI instructions — URLs to open, actions to perform, element selectors or ref IDs, data to extract, or screenshots to capture"
output: "Accessibility-tree snapshots with element refs, page text/HTML/attributes, screenshots (PNG/JPEG/PDF), JSON results, HAR recordings, visual diffs, React component trees, Web Vitals metrics"
synergy: []
upstream_repo: https://github.com/vercel-labs/agent-browser
original_source_url: https://github.com/vercel-labs/agent-browser/blob/main/README.md
original_author: Vercel Labs
security_audits:
gen_agent_trust_hub: pending
socket: pending
snyk: pending
house-style: exempt — machine artifact — browser automation, no prose output
---
# Agent Browser
A fast, native Rust CLI that gives AI agents full browser control via Chrome DevTools Protocol (CDP). Agents navigate pages, read accessibility-tree snapshots with compact element refs, click, fill, screenshot, manage tabs, handle auth, and run multi-step web workflows — all without Playwright or Node.js.
## Install
```bash
curl -fsSL https://agent-browser.sh | bash
```
Or via npm:
```bash
npx agent-browser@latest
```
## Load the skill prompt at runtime
```bash
agent-browser skills get core
```
This fetches the up-to-date workflow prompt directly from the source — avoids stale context.
## Key commands
```bash
# Open a page and get an accessibility snapshot
agent-browser navigate https://example.com
# Click an element by ref ID (from a prior snapshot)
agent-browser click @e3
# Fill a form field
agent-browser fill @e7 "hello@example.com"
# Take a screenshot
agent-browser screenshot
# Natural language control via AI Gateway
agent-browser chat "go to zynkr.ai and find the pricing section"
```
## Cloud providers
Supports local Chrome/Chromium, Browserbase, Browserless, Browser Use, Kernel, and AWS Bedrock AgentCore. Configure via environment variables or `~/.agent-browser/config.toml`.
## Observability
Live dashboard at `http://localhost:4848` — viewport, activity feed, and AI chat panel.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!