Google Turns the Web Agentic

Edition 157 | February 12, 2026

I tested this and couldn’t stop laughing. “Should I walk 50 meters to the car wash or drive?”

One model said walk. Claude said drive… because you need the car.

Common sense benchmark: passed.

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

In today’s issue…

• Google previews WebMCP, letting agents call websites like APIs

• T-Mobile offers live AI translation directly inside phone calls

• Google turns Search into an agentic storefront

• OpenAI ships Skills + Shell for stable long-running agents

• New tools to benchmark, secure, and stress-test your agents

…and more

🔥 INCASE YOU MISSED IT

Readers’ favorite items from the past week

1. OpenAI launches Frontier to manage AI agents like employees

2. NBCUniversal tests AI agent ad buying for live NFL games

3. Rentahuman.ai announces a marketplace for AI agents to hire real humans

4. GitHub adds Claude and Codex AI coding agents

5. Kenji Explains tests Claude in Excel, great for data cleaning and formula debugging

📌 THE THURSDAY BRIEFING

Google previews WebMCP for agent-website interactions

📡 Signal: Google's Chrome team released an early preview of WebMCP, a browser protocol that lets websites publish structured "tool contracts" so AI agents can call functions like buyTicket(destination, date) directly, instead of scraping screens or parsing the DOM.

For builders: This is potentially the biggest infrastructure shift for web-facing agents since MCP itself. Right now, browser agents waste tokens and time taking screenshots, guessing at page layouts, and simulating clicks. WebMCP flips that. The website tells the agent exactly what it can do and how to do it. If this gets adoption, it changes how you architect any agent that touches the web. Think of it as structured data for the agentic era, the same way schema markup made websites machine-readable for search engines, WebMCP makes them machine-actionable for agents.

Google makes Etsy and Wayfair shoppable in agentic search

📡 Signal: Google is letting AI agents browse, filter, and surface products from Etsy, Wayfair, and other retailers directly within its agentic search results, turning Google Search into a shopping agent that acts on your behalf.

For builders: This is the exact opposite of what eBay did a couple weeks ago when they banned AI agents from their platform entirely. Two very different bets on the same question: should agents be allowed to shop for people? Google is betting yes and building the infrastructure to make it seamless. The pattern to watch is whether more marketplaces follow Google's lead or eBay's, because which side a platform picks will determine whether your agent can access their inventory at all.

🤝 WITH WISPR FLOW

Vibe code with your voice

Vibe code by voice. Wispr Flow lets you dictate prompts, PRDs, bug reproductions, and code review notes directly in Cursor, Warp, or your editor of choice. Speak instructions and Flow will auto-tag file names, preserve variable names and inline identifiers, and format lists and steps for immediate pasting into GitHub, Jira, or Docs. That means less retyping, fewer copy and paste errors, and faster triage. Use voice to dictate prompts and directions inside Cursor or Warp and get developer-ready text with file name recognition and variable recognition built in. For deeper context and examples, see our Vibe Coding article on wisprflow.ai. Try Wispr Flow for engineers.

Try Wispr Flow

 🤖 AGENT OF THE WEEK

👋 Welcome back!

No Agent of the Week this edition, but I need your help deciding what comes next!

I've been building n8n agents for the last few months, and I love it. But the agentic landscape has exploded recently: Claude Code, Claude Co-Work, OpenClaw, browser agents, and I've been experimenting with a bunch of them behind the scenes (I even picked up a Mac Mini this week to test some of these locally).

So before I build the next one, I want to hear from you: what do you actually want to see?

Till next week,
✌️ AP