Any Website Can Now Have an AI Copilot

A small thing that changes a lot

Imagine your customers could type "show me all unpaid invoices from last month" into your software — and it just happened. No clicking through five menus. No training anyone. Just plain English, answered immediately.

That's what Alibaba's page-agent makes possible. It's a free, open-source tool that you (or your developer) can drop into any existing website or internal tool in minutes. Once it's there, users can describe what they want to do, and the software figures out how to do it on their behalf.

It works with the AI models you've probably already heard of — OpenAI, Claude, Gemini — and it doesn't require rebuilding anything from scratch. It sits on top of what you already have.

For businesses running complex internal systems — think ERPs, booking platforms, admin dashboards — this is quietly significant. The kind of thing that used to take a full engineering team to build is now something you can test in an afternoon.

It hit the top of GitHub's trending list this week, which usually means a lot of developers are paying attention. Worth keeping an eye on.

Words worth knowing

Open-source — The code is free and public. Anyone can use it, modify it, or build on top of it. It also means the community can catch bugs faster.

Script tag — A single line that developers paste into a website to add new functionality. Think of it like plugging in a toaster — you don't need to rewire the kitchen.

LLM (Large Language Model) — The brain behind tools like ChatGPT or Claude. Page-agent lets you choose which brain powers your copilot.

DOM — The live structure of a webpage. When the agent "executes against the DOM," it means it's actually clicking buttons and filling fields — not just talking about it.

---

If you have an internal tool your team finds frustrating to navigate, this is worth bringing up with your developer. The question to ask: could we add a search bar that actually understands what people mean?