A developer scanned 111,076 of the top 200,000 websites and found zero WebMCP implementations in production. Rather than waiting, they shipped WebMCP on two sites — a SvelteKit personal blog using the @mcp-b/global polyfill and a Next.js product using the .well-known/webmcp manifest approach. The post includes full implementation code for both approaches, adoption data showing a sharp cliff between passive AI signals (robots.txt, ai.txt) and active capability standards (WebMCP, A2A Agent Cards), and a practical argument for early adoption despite zero current agent traffic. Chrome 146 Canary has partial support; production browser support is estimated for 2027 at the earliest.
Table of contents
Table of ContentsPrerequisitesWhat WebMCP Actually Is (And Why It Matters in 2026)The Adoption Curve Nobody Is Talking AboutWhat I Actually Shipped: Two Sites, Two ApproachesWhat I Learned From Shipping Something Nobody Else HasThe Part That Actually Surprised Me: What the Adoption Curve Means for TodayHow to Ship WebMCP Today (Full Implementation Path)The Practical Answer to "Why Bother Now"Where This Goes NextSort: