The same word dropped from four major tech companies in the same April — "Skills". Anthropic Claude, OpenAI ChatGPT, Google Gemini, and on April 14th, Chrome itself.

Chrome adding Skills isn't just a feature update. It's a signal that AI skills have dropped down to the OS layer — with a real chance of becoming as universal as right-clicking. And in the same Chrome 146, there's an even bigger play: WebMCP (Web Model Context Protocol), now in beta — a new standard that lets websites expose their functions directly to AI agents.

What Is It?

On April 14, 2026, Google officially launched Skills in Chrome. You save AI instructions you use often (like "calculate the protein macros in this recipe" or "compare these two laptop specs"), then call them up instantly with a slash (/) or + shortcut the next time you're in the same situation. No more retyping the same prompt every time.

Here's the thing — Claude (Anthropic Skills), ChatGPT (Skills in ChatGPT), and Gemini (Agent Skills) all adopted the same pattern around the same time. "AI Skills" has become a category name, and Chrome just pulled that category down to the OS layer. Anthropic Claude's Skills format got the most traction on X — and now Google has baked that same pattern directly into the browser.

And there's an even bigger move inside the same Chrome 146 — WebMCP (Web Model Context Protocol). It's a new standard in dev trial that lets websites expose their actions directly to AI agents. Take flight booking as an example: before, an AI had to find the "Book a Flight" button on screen. With WebMCP, it can call bookFlight({origin, destination, date}) directly.

  1. Save a Skill
    Save a frequently-used prompt as a Skill in your Gemini in Chrome chat history. A ready-to-use Skills library is also available.
  2. Call it up
    Use the slash (/) or + shortcut in the address bar or side panel. Your saved Skills list appears instantly.
  3. Auto-inject context
    The content of the page you're currently viewing is automatically fed into the Skill as context.
  4. Confirm before running
    Skills that trigger an action ask for user confirmation first. Gemini in Chrome's safety guardrails apply in full.

Why This Actually Matters

The key is that this could be a signal of the end of "AI search engines". After the ChatGPT shock in 2023, everyone expected "AI browsers" like Perplexity and Comet to define the next era. But Google's answer is — quietly bake AI into the Chrome you already have, so there's no reason to switch to something new. It's the same playbook as AI Overviews, which is now generating billions of views in search results every month.

Approach Dedicated AI Browser (Comet·Atlas·Dia) Chrome 146 + Skills + WebMCP
Install friction Download a new browser, migrate everything Auto-update to Chrome you already have
User reach Millions of early adopters 3B+ Chrome users
Workflow relearning New tabs, extensions, shortcuts to re-learn Same Chrome you already know
Skills standard Each has its own format Effectively compatible pattern with Anthropic·OpenAI
Website-side integration Relies on scraping·UI automation Direct function calls via WebMCP

WebMCP is the genuinely interesting part. Search Engine Land's analysis calls this "the next evolution of SEO — agentic optimization beyond AEO." In B2B scenarios, quote requests, inventory checks, and logistics comparisons collapse into a single function call that hits multiple vendors simultaneously. In B2C, hotel, flight, and restaurant bookings can be integrated directly — no Expedia needed.

Getting Started

  1. Update to Chrome 146.0.7672.0 or later
    Regular users get Skills enabled automatically. WebMCP requires a flag.
  2. Enable Skills and browse the library
    Gemini in Chrome side panel → Skills tab → try 1–2 items from the built-in library.
  3. Build one Skill for your own workflow
    Pick something you do every week — clean up emails, sort receipts, summarize articles — and save it as a prompt.
  4. Experiment with WebMCP (developers)
    Enable chrome://flags/#enable-webmcp-testing → install the Model Context Tool Inspector extension → verify behavior on Google's demo site.
  5. Website owners: review only for now
    WebMCP isn't finalized yet, so hold off on production deployment. Do start learning the navigator.modelContext API.

FAQ

(Auto-rendered by FAQSection — not included in content)

Deep Dive Resources

The Deep View — Chrome's AI skills push the web toward agents April 14 launch coverage plus an Anthropic·OpenAI·Google Skills comparison thedeepview.com

Search Engine Land — WebMCP explained The difference between WebMCP's imperative and declarative APIs, B2B/B2C use cases, and an implementation guide searchengineland.com

Chrome for Developers — WebMCP early preview Official dev preview docs — spec details and testing procedures developer.chrome.com

bug0.com — WebMCP just landed in Chrome 146 A hands-on code guide for registering an AI tool from a single frontend JS file bug0.com