SEO made websites discoverable by search engines. WebMCP does the same for AI agents.
As AI agents move from chat windows into browsers – Chrome’s WebMCP Early Preview Program, Claude, Gemini, and others – websites need a standardized way to tell agents what they can do, not just what they contain. Without it, your site is invisible to the next generation of user interfaces. WebMCP is the discovery protocol that bridges that gap: a /.well-known/mcp-manifest.json endpoint that lets any agent understand your site’s tools, resources, and
brand voice in a single request.
Now imagine the business impact. A visitor’s AI agent lands on your product page and instantly knows it can add items to a cart, submit a contact form, or book an appointment – without scraping DOM elements or guessing at URLs.
Your website becomes a sales and marketing partner that speaks directly to agents, converting AI-assisted browsing into real transactions. Every content type you publish can either inform (resources) or act (tools), and your
brand voice travels with every interaction.
uWebMCP is an open-source Umbraco 17 package that makes this real today. It auto-generates MCP tool definitions from your content types and Umbraco Forms, exposes structured resources for blogs and products, enforces validation
to guard against hallucinations, and includes a backoffice “Agent Intelligence” dashboard to track which AI agents are discovering and invoking your site. Custom tools with real execution logic are supported via a simple
IWebMcpTool interface – define a “Buy Product” or “Book Appointment” tool in a few lines of C#.
Built on .NET 10, enterprise-ready with rate limiting, CSRF protection, honeypot spam filtering, and distributed caching out of the box.
The package is early, the code is public, and I’d genuinely appreciate feedback from the Umbraco community. Whether it’s architecture, use cases, or just “this is interesting” – I want to hear it.
