Angular WebMCP is Currently a Localhost Experience Only!

You might have heard about the WebMCP Angular support introduced in v22 through various blog posts explaining how to implement it in your Angular projects.

If you are looking for a deep dive into the concept, we highly recommend checking out this great article by Paweł Kubiak about teaching AI agents to interact with your web app.

But while the implementation side is well documented, many developers find themselves stuck when it comes to actually testing this feature in the wild.

The core reason? WebMCP is currently a very experimental feature in Chrome, and the ways to interact with it remain limited.

The Two Testing Paths

Currently, there are only two ways to interact with WebMCP locally:

1. The window.modelContext API

You can call the window.modelContext API (previously exposed by the navigator) to get the tools list and to execute a tool.

However, this won't place you in a real-world situation. The whole point of WebMCP is that you want an AI Agent to trigger these tools autonomously. If you are manually calling the tools via the API, you'd be better off just handling the logic directly in your application.

2. The WebMCP Web Extension

The other option is installing the WebMCP web extension.

While this sounds like a step closer to reality, it's not enough for a production environment. It requires activating specific WebMCP support flags in chrome://flags. This implies that the user knows exactly what they are doing and is comfortable tweaking internal browser settings. That's definitely not a behavior you should expect—or ask—an end user to adopt.

The Verdict for Production

What does this mean for your projects? You can absolutely give WebMCP a try and test it out locally to get familiar with the paradigm. But do not push it into production as a new user-facing feature yet. The ecosystem simply isn't ready for mainstream consumption.

What's Next?

The true potential of WebMCP will be unlocked with the upcoming browser and AI integration milestones. Specifically, we are waiting for:

• Prompt API Integration: Having the Prompt API natively support accessing WebMCP applications.
• Chrome Integration: The upcoming 'Ask to Gemini' Chrome integrated chat, which is expected to support seamless WebMCP interactions.

Until then, keep experimenting locally, prepare your agents, and stay tuned for the next evolution of AI interaction on the web.