Hey PH 👋 I built this over the weekend after reading the WebMCP spec and noticing the spec authors themselves flagged tool discoverability as an unsolved limitation. Quick background: WebMCP is a W3C proposal (currently in Chrome origin trial) that lets websites expose tools directly to AI assistants — natively in the browser, no backend integration required. A tool is just a named action your site supports: add-to-cart, search-flights, create-task. You define the name, a description, and an input schema. When a user visits your site with an AI assistant active, the assistant sees those tools and can invoke them directly instead of trying to click through your UI or scrape your HTML. The problem is discoverability. WebMCP tools are registered via JavaScript at runtime, they don't exist in HTML source. Crawlers can't find them. There's no way to know what tools a site exposes without actually loading and executing it. So agents have no way to know ahead of time what's out there. The only clean solution is an opt-in registry with domain verification. Site owners publish what they support, agents know where to look. That's what this is. You submit your domain, describe your tools, prove you own the domain with a DNS TXT record, and you're listed. Agents and developers can search the registry by domain or keyword and pull tool definitions via API before ever visiting a site. The read API requires no auth, it's meant to be queried freely. I think this is day one infrastructure for a meaningful shift in how AI agents interact with the web. Would love feedback from anyone building agents or thinking about exposing tools from their own site. https://webmcp-registry.dev

WebMCP Registry

The public index for WebMCP tool contracts

The public index for WebMCP tool contracts