page-agent

To embed an AI copilot, you integrate a client-side TypeScript package into your web page, which connects to an OpenAI-compatible LLM endpoint to interpret natural-language instructions and execute DOM operations like clicking and typing.

An in-page GUI agent is a client-side script that manipulates the DOM on a host page. For SaaS dashboards, it receives natural-language user instructions, uses an LLM to interpret them, and programmatically clicks, types, and navigates UI elements.

Yes, the in-page agent works with any OpenAI-compatible LLM endpoint. You configure the client-side PageAgent package to route natural-language instructions to your chosen endpoint, which then returns the DOM operations to execute on the host page.

You can add a natural-language navigation panel to legacy web apps by inserting a single script tag or installing the npm package. The PageAgent renders the panel client-side and executes DOM operations on the existing UI without requiring backend changes.

No, the web UI automation agent runs entirely client-side. It relies on the host page to render the panel and execute DOM operations, connecting directly to an OpenAI-compatible LLM endpoint without requiring a dedicated backend server.

You should avoid a client-side DOM manipulation agent if your web app requires server-side data processing, lacks accessible UI elements for interaction, or if exposing client-side LLM instructions poses security constraints for your specific environment.