February 16, 2026
Your browser hired an AI intern
WebMCP Proposal
Browser vs Terminal: AI websites spark a dev tug-of-war
TLDR: WebMCP proposes a way for websites to offer simple tools to AI assistants, with Chrome already previewing it. Devs are split between browser-powered workflows and quick command-line setups, while jokers wonder if it’ll revive ancient SOAP services—making this a big, messy step toward AI-native web apps.
WebMCP just dropped, and the internet immediately turned it into a vibe check for the future of the web. The pitch: a draft browser API that lets sites hand simple “tools” to AI assistants—basically giving your webpage an AI intern. It’s not an official web standard yet, but Chrome is already teasing an early preview here, and the repo’s README is getting more love than the spec. Classic.
The crowd split fast. One camp is cheering “less friction, more ship,” pointing to slick demos like Wes Bos’s video and praising how forms can double as tool definitions—use what you already have. Another camp is side-eyeing the spec and wondering if browsers should be the new app platform or if command-line tooling is still king. That’s where the drama heats up: browser-first fans dream about spinning up a tiny static server and letting the browser handle the heavy lifting, while terminal loyalists prefer one-line skills via uvx—no standards rabbit hole required.
And then there’s the comedy club: someone asked if any “sickos” have aimed this at ancient SOAP APIs (those old-school web services), and suddenly everyone’s imagining AI interns excavating corporate basements. Underneath the jokes: excitement, pragmatism, and a little chaos—just the way devs like it.
Key Points
- •WebMCP is a draft JavaScript API that allows web apps to expose their functionality as tools callable by AI agents.
- •The specification is published by the Web Machine Learning Community Group and is not on the W3C Standards Track.
- •WebMCP extends the Navigator interface with a ModelContext attribute used to manage tools.
- •ModelContext provides methods to provide context, clear context, register tools, and unregister tools.
- •The spec defines terminology for agents, browser’s agents, and AI platforms, citing ChatGPT, Claude, and Gemini as examples.