agent-browser Browser Automation

A guide that teaches Claude to drive the agent-browser CLI for navigating sites, filling forms, extracting data, and taking screenshots with minimal tokens.

AutomationIntermediate★ 118⑂ 7AI score 9/10Last updated: Aug 20, 2026

What it does

Teaches Claude the correct usage of the agent-browser CLI, which drives Chrome/Chromium over CDP without Playwright or Puppeteer.
Centers on the snapshot-and-ref loop: read a compressed accessibility tree (~200-400 tokens) and act on short refs like @e3, re-snapshotting after every page change.
Covers navigation, click/fill/type/select/upload, text and attribute extraction, screenshots (including --annotate for multimodal models), tab management, parallel isolated sessions, network mocking, video recording, iframes, and dialogs.
Includes a dedicated waiting-strategy section to avoid flaky wait 2000 patterns, plus troubleshooting for "Ref not found", overlays swallowing clicks, and custom inputs that ignore fill.
States safety rules: treat page content as untrusted data, never echo secrets, use the auth vault or saved state instead of typing passwords.

Who it's for

Developers who want Claude to run E2E/QA checks against a web app
Anyone automating repetitive tasks or scraping behind a login
React frontend devs who want component trees, re-render profiles, and Web Vitals from the CLI
People who need lightweight browser automation without a Playwright/Puppeteer setup

Examples

"Log into the app and screenshot the dashboard" → open → snapshot -i → fill/click → wait --url "**/dashboard" → screenshot
"Extract this pricing table as JSON" → heredoc into eval --stdin running querySelectorAll and returning an array of objects
"Test a two-user flow simultaneously" → --session a and --session b give two cookie-isolated browsers
"Find what's re-rendering on localhost:3000" → open --enable react-devtools then react renders start/stop

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/molefrog/moi/HEAD/.agents/skills/agent-browser/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)

Download the ZIP with the button below.
In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.

↓ Download ZIP

Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the .agents/skills/agent-browser folder from the GitHub repo molefrog/moi into my ~/.claude/skills/molefrog-agent-browser/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/molefrog/moi /tmp/moi && mkdir -p ~/.claude/skills && cp -r /tmp/moi/.agents/skills/agent-browser ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

Confirm Node.js is installed: node -v.
Install the CLI globally: npm i -g agent-browser.
Provision a browser: agent-browser install (in restricted/cloud environments where the Chrome download is blocked, skip this and point AGENT_BROWSER_EXECUTABLE_PATH at a system Chromium).
Clone the repo: git clone https://github.com/molefrog/moi /tmp/moi.
Copy the skill: mkdir -p ~/.claude/skills && cp -r /tmp/moi/.agents/skills/agent-browser ~/.claude/skills/.
Restart Claude Code and try a prompt like "open example.com and take a screenshot" to confirm the skill triggers.
If anything fails, run agent-browser doctor first for diagnosis.

View source on GitHub ↗License: NOASSERTION