Agentic Browser for Openclaw

A sophisticated browser automation tool that enables AI agents to navigate the web, interact with elements, and record sessions using the inference.sh infrastructure.

xyny89
v1.0.0
Feb 10, 2026
0
2.5k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install agentic-browser-0-1-2

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install agentic-browser-0-1-2 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 Agentic Browser?

Agentic Browser is a powerful automation skill designed to give AI agents human-like browsing capabilities. By leveraging the Playwright framework through inference.sh, it allows agents to navigate complex web applications, interact with UI components using a simplified @e reference system, and execute custom JavaScript.

This skill is an essential part of the Openclaw Skills library, providing the necessary bridge for agents to perform live web research, automate repetitive form-based tasks, and capture visual evidence through screenshots and video recordings. It handles the heavy lifting of session management and element state tracking, making web-based agentic workflows more reliable and easier to implement.

Agentic Browser Use Cases

  • Automating multi-step form submissions and account registrations
  • Extracting real-time data from dynamic websites for research purposes
  • Performing automated UI/UX testing and regression checks
  • Generating video tutorials or documentation of web-based workflows
  • Navigating behind authenticated portals for data gathering

How Agentic Browser Works

  1. Start a new session by navigating to a target URL, which initializes the browser environment and generates @e element references.
  2. Interact with the page by sending commands like click, fill, or scroll using the specific element references provided by the initial snapshot.
  3. Refresh the page state using the snapshot function after any significant DOM changes or navigation to ensure element refs remain valid.
  4. Optionally capture visual data through full-page screenshots or by enabling high-fidelity video recording during the session.
  5. Terminate the session to clean up resources and retrieve any generated media files or final data payloads.

Agentic Browser Setup

To get started with this component of Openclaw Skills, you need to install the inference.sh CLI and authenticate your environment:

# Install the CLI tool
curl -fsSL https://cli.inference.sh | sh && infsh login

# Run a test command to open a website
infsh app run agent-browser --function open --input '{"url": "https://example.com"}' --session new

Agentic Browser Data Schema & Taxonomy

The skill organizes web interactions through structured sessions and a reference-based element system.

Data Type Description
@e Refs Unique identifiers for interactive elements (e.g., @e1 for a button)
Session ID A UUID used to maintain state across multiple functional calls
Screenshots Viewport or full-page image files generated on demand
Video Files MP4 recordings of the browser session provided upon closing
JS Results Raw JSON or string output from custom JavaScript execution strings

Agentic Browser Advanced Features

  • Visual cursor indicators in videos and screenshots to track agent movements
  • Full proxy server support with authentication for geo-specific browsing
  • Drag-and-drop support for complex web interfaces and file uploads
  • Persistent session state allowing for authenticated login reuse across tasks
  • Deep integration with the Openclaw Skills ecosystem for chaining research and analysis tasks

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*