Julien Reszka
WebMCP is making agentic AI viable directly in the browser and it is giving me a strong feeling of deja vu.
In 2015 I was converting old desktop websites to mobile responsive because the shift to mobile had made them broken and unusable. Most businesses had not seen it coming. The technical shift happened, the user behavior followed, and suddenly there was a wide-open gap between what existed and what was needed.
WebMCP is doing something structurally similar. It gives browsers a standard way to connect AI agents to tools and data sources without routing everything through a backend server. Agentic AI that browses, fills forms, extracts data, and takes actions can now live entirely in the client. The infrastructure assumption that AI requires a backend is starting to break the same way the assumption that web apps required a desktop did.
What this opens up:
- Privacy-first AI tools that never send user data to a server
- Offline-capable agents that work without a network connection
- Dramatically lower operating costs for AI-powered products
- A new class of browser extensions that are actually intelligent
- Personalized models that run and adapt on the user's own hardware
In 2015 the businesses that moved early on responsive design built a durable advantage. The ones that waited paid someone like me to fix it under pressure later.
I do not know exactly how the agentic browser shift will play out. But I know what it feels like when a platform constraint disappears and the old assumptions stop being true. This feels like that.
Discussion