A powerful browser automation skill for AI agents that connects to real Google Chrome instances via the Chrome DevTools Protocol (CDP).
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install chrome-for-openclaw
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install chrome-for-openclaw using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Chrome for OpenClaw?
Chrome for OpenClaw is a specialized browser automation CLI designed for AI agents. Unlike standard headless browsers, this Openclaw Skills integration connects directly to a running Google Chrome instance via the Chrome DevTools Protocol (CDP). This allows AI agents to leverage existing user login sessions for services like Gmail or GitHub, effectively bypassing complex authentication hurdles.
By using the agent-browser CLI, agents can navigate pages, fill forms, extract data via interactive snapshots, and capture screenshots. The skill is optimized for environments using XRDP and XFCE, ensuring a stable and persistent browser environment for long-running automation tasks.
Chrome for OpenClaw Use Cases
- Automating tasks within authenticated web sessions like Gmail, GitHub, or internal dashboards.
- Web scraping and data extraction from complex, JavaScript-heavy websites.
- Automated form filling and multi-step transaction processing.
- Generating annotated screenshots and PDFs for reporting or visual debugging.
- Parallel browser session management for multi-agent workflows.
How Chrome for OpenClaw Works
- Initialize Connection: The skill connects to a Chrome instance running with the --remote-debugging-port=9222 flag.
- Navigation: The AI agent issues an 'open' command to move the browser to a specific URL.
- DOM Snapshotting: An interactive snapshot is taken to generate unique element references (e.g., @e1, @e2) for the current page state.
- Action Execution: The agent uses these references to click, type, or scroll, or uses semantic locators (text, label, role).
- State Management: Authentication and session data are persisted naturally within the real Chrome profile or exported to JSON for reuse.
Chrome for OpenClaw Setup
Install and configure the environment using the following steps:
# Step 1: Install Chrome, XRDP, and XFCE dependencies
bash <(curl -fsSL https://raw.githubusercontent.com/joustonhuang/chrome_for_openclaw/main/chrome_for_openclaw.sh) --install
# Step 2: Launch Chrome in CDP debug mode
bash <(curl -fsSL https://raw.githubusercontent.com/joustonhuang/chrome_for_openclaw/main/chrome_for_openclaw.sh)
# Step 3: Install the agent-browser CLI globally via npm
npm install -g agent-browser
# Step 4: Export the CDP URL for seamless connectivity
export AGENT_BROWSER_CDP_URL=http://127.0.0.1:9222
Chrome for OpenClaw Data Schema & Taxonomy
The skill manages browser interactions and data through several mechanisms:
| Data Type | Description |
|---|---|
| Interactive Refs | Temporary pointers (@e1, @e2) mapping to DOM elements within a specific snapshot lifecycle. |
| Session State | JSON files (e.g., auth.json) containing exported cookies and localStorage for session persistence. |
| Artifacts | Screenshots (PNG), annotated images, and PDF exports generated during automation. |
| Network Logs | Captured XHR and Fetch request data available via the network command. |
Chrome for OpenClaw Advanced Features
- Batch Execution: Chain multiple commands (open, wait, click, snapshot) into a single execution to reduce latency.
- Session Multiplexing: Run and manage multiple parallel browser sessions using the --session flag.
- Vision Mode: Generate annotated screenshots that provide a visual legend for the AI to better understand page layout.
- Observability Dashboard: Launch a live dashboard on port 4848 to monitor agent browser actions in real-time.
- Security Sandboxing: Implement content boundaries and domain allow-lists to protect against prompt injection and unauthorized navigation.
SKILL.md
Loading
Related Openclaw Skills
Kling AI Image to Video
An advanced AI-driven animation skill that converts static images into professional 1080p video clips using natural language commands.
Best Video Editing Software
This skill automates professional video editing and cloud rendering, allowing you to convert raw footage into polished 1080p MP4 files using AI.
Candy AI
Candy AI is a high-performance cloud-based video generation tool that transforms simple selfie clips and images into interactive 1080p AI companion character videos.
Free Video Generation by AI
A powerful AI video creation tool that transforms text descriptions into high-definition 1080p MP4 videos using cloud GPU rendering.
Video Editing Tools
An AI-powered video editing skill that automates trimming, transitions, and 1080p exporting via cloud GPUs.
Legal/TOS Diff-er
A specialized AI agent skill designed to track, snapshot, and semantically analyze changes in Terms of Service and legal documents.
Featured*
