agent-browser

Automate browser interactions via CLI with ref-based selectors.

Updated Jan 21, 2026

Search Tags:#screenshots #cli #bash #web-scraping #browser-automation #ref-based-selectors

Ready to run

npx skills add https://github.com/ollieb89/orchestrator --skill agent-browser-ollieb89

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-browser
Source: https://github.com/ollieb89/orchestrator/tree/main/.cursor/skills/agent-browser
Command: npx skills add https://github.com/ollieb89/orchestrator --skill agent-browser-ollieb89

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Automates browser interactions for AI workflows by providing a CLI-based, ref-driven tool that replaces more heavy-weight MCPs, enabling tasks like form filling, clicking, and data extraction without manual scripting.

Core Features & Use Cases

Ref-based interactions: execute actions using stable element references (e.g., @e1, @e2) to minimize brittle selectors.
Headless and headed modes: run in environments with or without a visible browser.
Screenshots and inspection: capture screenshots or inspect page structure for validation or debugging.
Use cases: automate login flows, perform quick web-assisted data gathering, or test simple automation pipelines from the command line.

Quick Start

Ask the agent to open a URL and perform a basic automation task using the agent-browser CLI.

Frequently Asked Questions about agent-browser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I automate browser interactions from the command line for an AI agent?▼

Browser automation for AI agents can be handled via a CLI tool that executes commands like open, snapshot, click, and fill. It enables navigation, element interaction, and data capture directly from the command line.

How do I use ref-based selectors for web scraping instead of brittle CSS selectors?▼

Ref-based selectors replace brittle CSS selectors by using stable element references like @e1 or @e2. You execute a snapshot command to generate these references, then pass them to click or fill commands for reliable web scraping.

What is the best way to take screenshots during web scraping and browser automation?▼

Taking screenshots during web scraping is done via a dedicated screenshot command in the CLI. This captures the current page state for validation, debugging, or data capture during headless or headed browser automation sessions.

Can I run CLI browser automation in headless mode without a visible browser window?▼

Yes, CLI browser automation supports both headless and headed modes. You can run sessions in environments without a visible browser, making it suitable for automated login flows and data gathering on servers.

Does this browser automation CLI replace heavy-weight MCPs for form filling tasks?▼

This CLI-based, ref-driven tool replaces heavy-weight MCPs by enabling tasks like form filling, clicking, and data extraction without manual scripting. It provides a lightweight alternative for quick web access workflows.

What are the limitations of using a CLI for browser automation and data extraction?▼

The CLI requires the agent-browser tool to be installed locally and relies on ref-based selectors generated from snapshots. It is designed for quick web access and simple automation pipelines, not heavy-weight MCPs.