page-agent

To embed an AI copilot inside your web app, install the page-agent package via CDN or npm and configure the PageAgent API with your model, baseURL, and language. This adds a client-side assistant that reads the DOM and executes natural language instructions directly in the browser.

Yes, you can drive UI interactions without server-side browser automation by using a client-side AI copilot. It interprets the DOM as text and executes commands like "click login" or "fill username" entirely within the browser, eliminating the need for backend calls.

Yes, the in-page AI copilot works with legacy web apps, SaaS dashboards, and admin panels. It ships as a pure TypeScript single script or npm package, allowing you to easily embed the assistant into various web environments to evaluate UI flows or enhance accessibility.

To set up a client-side AI agent, you need a modern browser and a compatible OpenAI-like LLM endpoint. You must configure the PageAgent API with specific parameters including the model, baseURL, and language to enable the copilot to process natural language instructions.

The best way to add an AI assistant for accessibility is embedding a client-side copilot that reads the DOM as text. This allows users to drive page interactions through natural language commands, enabling accessibility tools directly without server-side automation.

Running an AI copilot entirely in the browser requires a modern browser and a compatible OpenAI-like LLM endpoint to function. It operates purely client-side by reading the DOM as text, which means it depends on the browser environment and the external LLM configuration provided through the PageAgent API.