- Home
- Skills
- Agents & Orchestration
- agent browser
agent browser
A browser automation CLI for AI agents to navigate, interact, and extract data from the web using stable @refs.
$1.49
Works with the AI Agents you already use
agent browser
Example session with this skill installed
Open https://news.ycombinator.com, take a snapshot of the top stories, and then take a screenshot.
- Read your context and instructions
- Compiled the agent browser
I have navigated to Hacker News, captured the interactive element references for the top stories, and saved a screenshot of the current page layout.
Browser Automation Result
| Action | Result | Details |
|---|---|---|
| Navigate | Success | Loaded https://news.ycombinator.com |
| Snapshot | Captured | 30 story links identified (@e1-@e30) |
| Screenshot | Saved | /tmp/agent-browser/screenshot_1715684201.png |
Identified Elements (Sample)
@e1: [link] "Why I'm sued by a patent troll"@e2: [link] "Show HN: A new way to build agents"@e3: [link] "The history of Lisp"@e4: [button] "More"
Next steps
- Use
agent-browser click @e1to navigate to the first story. - Run
agent-browser get text @e1to extract the specific headline text. - Use
agent-browser batchto click a story and capture its content in one call.
Connects securely to your tools. The creator never sees your data.
About this skill
Building browser-based agents often fails due to complex DOM structures, flaky selectors, and authentication hurdles. This skill provides a CLI-driven browser automation layer designed specifically for LLM agents to interact with the web using Chrome/Chromium via CDP.
What it does
- Smart Element Discovery provides an accessibility tree snapshot with short @refs for easy element targeting.
- Session Management maintains state across browser restarts using sessions, persistent profiles, or auth vaults.
- Visual Verification captures screenshots, annotated views with element labels, and visual diffs to verify actions.
- Network Control inspects, filters, or blocks XHR/Fetch requests and records HAR files for debugging.
- Multi-Platform Support executes automations on local Chrome, iOS Simulators, or cloud providers like AgentCore and Browserbase.
- Efficiency Tooling supports command chaining and batching to minimize tool calls and context window usage.
How it works
- Target Discovery: Run
agent-browser openfollowed bysnapshot -ito see interactive elements and their assigned refs. - Interaction: Execute commands like
click @e1orfill @e2using the discovered refs to manipulate the page. - Observation: Use
screenshotorget textto extract data and verify the page state after interaction. - Automation: Batch multiple commands together using the
batchcommand to perform complex sequences in a single call.
Frameworks & tools
Built for Chrome/Chromium via Chrome DevTools Protocol (CDP). Supports Node.js, Homebrew, and Cargo installations. Compatible with iOS Simulators (Appium/Xcode) and cloud browser providers including AgentCore, Browserbase, and Lightpanda.
Why this beats prompting it yourself
Writing raw Puppeteer or Playwright code inside a prompt is brittle and consumes excessive tokens. This skill abstracts the browser into a high-level CLI that uses stable @refs, handling frame inlining and session persistence automatically so the agent stays focused on the logic rather than the implementation details.
Use cases
- Automate complex form submissions across multiple pages with persistent login state.
- Scrape dynamic web content by discovering elements semantically and extracting clean text or JSON.
- Perform visual regression testing by comparing snapshots or screenshots against a baseline.
- Build cross-platform agents that run identical automation scripts on both desktop and mobile layouts.
Known limitations
Lightpanda engine does not support extensions, profiles, or state files. iOS automation requires a macOS host with Xcode and Appium installed. Default timeouts may need adjustment for high-latency sites.
Lightpanda engine lacks extension/profile support. iOS requires macOS/Xcode/Appium. Networkidle wait strategy may hang on sites with persistent tracking or websockets.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 8 days ago
Requires
Permissions
Allowed Hosts