A high-performance Rust-based headless browser CLI designed for AI agents to interact with the web via structured commands.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install miu
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 miu 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 command-line interface that bridges the gap between AI agents and the web. Built with a fast Rust-based core and a reliable Node.js fallback, it allows developers to integrate complex web navigation and data extraction into their Openclaw Skills. It treats web pages as structured environments where agents can identify interactive elements through stability-focused references.
By leveraging the accessibility tree and semantic locators, Agent Browser ensures that AI models can "see" and manipulate web interfaces with high precision. Whether you are building automated testers, data scrapers, or autonomous assistants, this tool provides the low-level control and high-level abstractions necessary for modern web-based AI workflows.
Agent Browser Use Cases
- Automating complex multi-step web interactions and form submissions.
- Extracting structured data and accessibility trees for AI processing.
- Programmatic UI testing and cross-device emulation for web applications.
- Building autonomous agents that require session persistence and authentication bypass.
- Creating high-fidelity video recordings and traces of automated browser sessions.
How Agent Browser Works
- Initialization: The agent starts a browser session using a specific URL or a saved state.
- State Analysis: A snapshot is taken (typically using the
-iflag) to generate a map of interactive elements and their unique @refs. - Interaction: The agent executes commands like
click,fill, orhoverby targeting the @refs identified in the snapshot. - Dynamic Update: After navigation or DOM changes, the agent re-snapshots the page to refresh references and ensure accuracy.
- Data Export: Results are captured through screenshots, PDFs, or JSON-formatted output for further analysis.
Agent Browser Setup
Install the CLI globally and ensure dependencies are met for your environment:
### Build from Source
```bash
git clone [https://github.com/vercel-labs/agent-browser](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 its interactions using a session-based architecture. Key data artifacts include:
| Component | Description |
|---|---|
| Snapshots | Hierarchical accessibility trees or compact interactive element lists returned as text or JSON. |
| Session State | JSON files containing cookies and localStorage to persist authentication. |
| Media | PNG screenshots (standard or full-page), PDF exports, and WEBM video recordings. |
| Refs | Volatile identifiers (e.g., @e1) that map CLI commands to specific DOM elements during a session. |
Agent Browser Advanced Features
- Session Persistence: Save and load full authentication states (cookies/storage) to bypass login flows in subsequent runs.
- Parallel Sessions: Run multiple isolated browser instances simultaneously using the --session flag.
- Network Orchestration: Intercept, block, or mock network requests directly from the CLI to simulate various edge cases.
- Device Emulation: Spoof viewports, user agents, and geolocations to test responsive designs or region-locked content.
- CDP Integration: Connect to existing browser instances via the Chrome DevTools Protocol for advanced debugging.
SKILL.md
Loading
Related Openclaw Skills
Strategic Decision Making
A high-stakes executive-mode skill for diagnosing complex problems and designing definitive strategies with four distinct rigor modes.
Bitbucket Cloud (Read-Only)
A secure, read-only bash CLI wrapper for AI agents to browse Bitbucket repositories, pull requests, and source code.
Ecommerce Product Pro
An AI-driven research suite for Amazon FBA, Shopify, and dropshipping designed to identify winning products and analyze market competition.
Retently AI Agent Skill
A powerful integration for AI agents to manage Retently surveys, customer sentiment, and feedback data through the Membrane framework.
CLIProxy + OpenClaw
A bridge that transforms subscription-based CLI and OAuth accounts into a unified, OpenAI-compatible API layer for seamless integration with AI agents.
Vector Databases (Deep Workflow)
A professional deep-dive workflow for selecting, optimizing, and operating vector databases for high-performance AI applications.
Featured*
