Breaking

Google Unveils WebMCP and Chrome AI Agent Tools at I/O 2026

June 16, 2026 at 18:22 EDT

AI Agents
Foundation Models

At its developer conference Google I/O 2026, Google announced updates to the AI platform for its Chrome browser, including WebMCP, a proposed standard that lets websites expose tools to AI agents; Built-in AI, on-device models; and Skills in Chrome, which lets users reuse prompts in Gemini in Chrome.

Google I/O 2026 · Chrome

Chrome Opens the Browser to AI Agents

Google unveiled WebMCP, Built-in AI, and Skills in Chrome — browser-level capabilities for an "agentic web" where AI agents operate websites autonomously, rather than just chat.

90%

Token usage cut vs Playwright loops in reported cases

149

Chrome version starting the WebMCP Origin Trial

Total Chrome updates announced at I/O 2026

Token Cost per Agent Task

WebMCP exposes executable tools directly, instead of guessing the UI via screenshots and DOM scraping.

100%

Playwright
loop

→

~10%

WebMCP
tools

Three Browser-Level Capabilities

WebMCP

Proposed open standard exposing structured tools (JS functions, forms) to agents. Origin Trial in Chrome 149 → shipping expected in 157.

Built-in AI

Server-free, on-device AI — translation, summarization, writing, rewriting running locally inside the browser.

Skills in Chrome

Save and reuse Gemini prompts via one click or a / command across tabs. Rolling out on Mac, Windows, ChromeOS (English US).

Why developers like it

Deterministic actions — no guessing the UI
Sharp drop in token usage and cost
Streamlines repetitive tab and research tasks

Open questions

API shape may change at Origin Trial stage
Agent hijacking and security risks
Needs added context: policy, state, exceptions

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.