TL;DR
A lightweight JavaScript agent framework that controls web interfaces using natural language within a webpage.
Key features
Easy integration: Operates with just a script tag injection, no browser extensions, Python, or headless browsers needed.
Text-based DOM manipulation: Directly interprets and manipulates the DOM structure without screenshots or multimodal LLMs.
Bring your own LLMs: Connect and use the language model of your choice.
Multi-page extension: Perform tasks across multiple tabs using a Chrome extension.
MCP Server support: Provides an interface for external agent clients to control the browser.
When to use it
When you want to quickly add AI copilot functionality to a SaaS product.
When you need to simplify complex web form inputs in ERP, CRM, or admin systems using natural language.
When you need to enhance web application accessibility by supporting voice commands, etc.
When you need to implement automation tasks spanning multiple web pages.