A high-performance Rust-based headless browser automation CLI that enables AI agents to navigate, interact with, and extract data from web pages.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-browser-bak
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 agent-browser-bak 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 Agent Browser?
Agent Browser is a robust tool designed to empower AI agents with the ability to navigate the web like a human user. By providing a fast, Rust-based headless browser interface with a Node.js fallback, it allows agents to execute structured commands for clicking, typing, and scrolling. This skill is essential for developers looking to expand their Openclaw Skills by giving agents a reliable way to interact with complex web applications and retrieve real-time information.
The core strength of Agent Browser lies in its ability to translate the complex Document Object Model (DOM) into a simplified accessibility tree. This enables AI agents to identify interactive elements through stable reference IDs, making web automation more resilient and easier to program than traditional scraping methods. Whether you are building research bots or automated testers, adding this to your library of Openclaw Skills provides the necessary bridge between LLMs and the live web.
Agent Browser Use Cases
- Automating complex multi-step form submissions and web UI interactions.
- Extracting structured data from JavaScript-heavy websites using accessibility snapshots.
- Performing automated cross-browser testing and visual regression checks.
- Creating web-based research agents that can navigate and summarize online content.
- Managing authenticated sessions by saving and loading browser states within the Openclaw Skills framework.
How Agent Browser Works
- Navigation: The agent opens a target URL to initialize a browser context and load the web content.
- Snapshotted Analysis: The skill generates an interactive snapshot, assigning unique reference IDs (e.g., @e1, @e2) to clickable and input elements.
- Interaction Execution: The agent sends commands like click, fill, or scroll using the specific reference IDs generated in the previous step.
- Dynamic Waiting: The tool can wait for specific network states, elements, or JavaScript conditions to ensure the page is ready for the next action.
- State Persistence: Session data including cookies and local storage can be saved to JSON files for future authenticated access.
Agent Browser Setup
Install the package globally using npm and set up the necessary browser dependencies:
npm install -g agent-browser
agent-browser install
agent-browser install --with-deps
For developers building Openclaw Skills from source:
git clone https://github.com/vercel-labs/agent-browser
cd agent-browser
pnpm install
pnpm build
agent-browser install
Agent Browser Data Schema & Taxonomy
Agent Browser organizes interaction data and page metadata to be easily consumable by AI models:
| Data Type | Description | Output Format |
|---|---|---|
| Snapshots | Accessibility tree of the page with interactive refs | Text/JSON |
| Session State | Cookies, local storage, and session tokens | JSON File |
| Visuals | Full-page or element-specific captures | PNG/PDF |
| Recordings | Video logs of the interaction sequence | WebM |
| Network Logs | List of tracked or intercepted requests | JSON |
Agent Browser Advanced Features
- Parallel Session Management: Run multiple isolated browser instances simultaneously using the --session flag.
- Network Mocking and Routing: Intercept, block, or mock network requests to simulate specific server behaviors.
- Semantic Locators: Find elements using natural language roles, labels, or text content instead of fragile CSS selectors.
- Headed Debugging Mode: View the actual browser window in real-time to troubleshoot complex Openclaw Skills workflows.
- CDP Integration: Directly connect to the browser via the Chrome DevTools Protocol for low-level manipulation.
SKILL.md
Loading
Related Openclaw Skills
MailCheck Email Verification
A powerful email validation skill for verifying addresses, bulk processing, and detecting phishing or fraud via the MailCheck API.
Apify
Run and manage over 17,000 specialized web scrapers and automation tools via the Apify REST API.
Azure Infra
A chat-based Azure infrastructure assistant utilizing the Azure CLI for resource querying, auditing, and safe management.
AWS Infra
A chat-based AWS infrastructure assistant that enables secure resource querying, auditing, and monitoring via the local AWS CLI.
EvoMap Auto Task Publish
An automated task executor for EvoMap that handles the full lifecycle of fetching, claiming, and publishing tasks to earn credits and reputation.
EvoMap Lite Client
A feature-rich lightweight client for AI agents to participate in the EvoMap ecosystem through automated task loops and Swarm collaboration.
Featured*
