I Built a Bridge for Google's New WebMCP Draft Spec — Here's What Broke
Hi, everyone. I usually go by tanahiro2010 in Japan. I'm a member of GDG Greater Kwansai. I gave the first-half talk at the Google I/O Extended Osaka 2026 hands-on session, "Let's Build WebMCP and Call It from an AI Agent!" This is a write-up of that hands-on session for Qiita (translated here for Dev.to). Here's the codelab we used: https://learn.gdgs.jp/webmcp-agent/ This is for people who…
Dear reader, I am Tanahiro2010, a member of GDG Greater Kwansai who recently presented a hands-on session on WebMCP at Google I/O Extended Osaka 2026. In this article, I will share my experience with the new WebMCP spec, which has some unique features that set it apart from the traditional Model Context Protocol (MCP).
MCP is a common interface for connecting external tools to an AI Agent, allowing tools to be registered and called throughout the Agent's session. In contrast, WebMCP is a draft spec published by the W3C Web Machine Learning Community Group, which mainly focuses on browser-embedded agents and tools. The spec allows users to register tools via JavaScript or by using a declarative form.
One of the key differences between MCP and WebMCP is the session lifetime. While MCP tools remain available as long as the Agent is running, WebMCP tools are only available while the page is open. This means that switching tabs will cause the tool to disappear from view. Despite this difference, the API structure of the two specs is quite similar, making it easy to transition between them.
As of August 2026, WebMCP is still in the proposal stage and not yet widely available. However, Chrome has introduced an Origin Trial for the feature, which can be enabled through chrome://flags/#enable-webmcp-testing. It is important to note that the API may still change during the draft phase, and the trial is not guaranteed to be available across all browsers.
In my hands-on session, I built a WebMCP Bridge to address the gap between the spec requirements and the actual implementation. This bridge allows users to connect external tools to an AI Agent using WebMCP, even though the spec is still in development. The bridge serves as a proof-of-concept that could potentially pave the way for future integration of WebMCP into AI Agent workflows.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
