The Problem

AI agents are powerful, but they're locked out of most websites. Custom integrations for every site don't scale. Each agent needs custom code to understand a particular website's structure, forms, and interaction patterns.

This creates a world where:

Agents miss 99% of the web
Building integrations is manual and expensive
Site updates break agent access instantly
There's no standard way for sites to expose capabilities to agents

WebMCP: The Solution

WebMCP is an open protocol that lets any website publish a machine-readable manifest of what an AI agent can do with it. Instead of custom integration code, agents read the manifest and adapt automatically.

Think of it as RSS feeds for AI agents: a standardized format that websites publish once, and agents use universally.

Without WebMCP

1. Agent developer writes custom scraper for site.com

2. Site changes design or adds CSRF tokens

3. Scraper breaks; agent goes offline

4. Developer writes new scraper for next site

5. Repeat 1,000 times for 1,000 sites

With WebMCP

1. Site publishes .webmcp manifest

2. Agent reads manifest, understands capabilities

3. Site updates design; updates manifest

4. Agent re-reads manifest, continues working

5. One protocol, infinite sites

How It Works

Every website publishes a WebMCP manifest at /.webmcp.json:

{
  "version": "1.0",
  "actions": [
    {
      "name": "search",
      "method": "POST",
      "endpoint": "/search",
      "fields": [
        {"name": "q", "type": "text", "required": true}
      ]
    },
    {
      "name": "book_appointment",
      "method": "POST",
      "endpoint": "/appointments/create",
      "fields": [
        {"name": "date", "type": "date"},
        {"name": "time", "type": "time"},
        {"name": "name", "type": "text"}
      ]
    }
  ]
}

Agents fetch this manifest once, then interact with the site using declared actions. When the site updates its capabilities, it updates the manifest. Agents adapt automatically.

Why This Matters

Speed to integration: Agents integrate with sites in minutes, not weeks
Maintenance-free: Site updates don't break agent access when manifest is kept current
Declarative: Sites describe what they offer; agents decide how to use it
Standardized: One protocol, thousand of agents, millions of sites
Open: Not owned by any AI vendor; a public standard anyone can implement
Secure: Sites can require authentication, rate limiting, and audit logging

Use Cases

E-commerce

Agents browse your store, fetch product details, create orders, check status - all via WebMCP endpoints you control.

Service Booking

Publish your calendar, availability, pricing, and booking process in a manifest. Agents schedule appointments on your behalf.

Content Management

Agents interact with your CMS via WebMCP: search posts, read metadata, understand taxonomy, publish on your site.

Lead Generation

Agents fill out your contact forms reliably with WebMCP endpoints; you get cleaner data, fewer parsing errors.

Getting Started

Publishing WebMCP is simple:

Identify the key actions agents should be able to perform on your site
Create a JSON manifest describing those actions
Host the manifest at /.webmcp.json
Agents read your manifest and start integrating automatically

For full specification, tools, and examples, visit the WebMCP documentation.

Build the AI-Native Web

WebMCP is open-source and open to contributions. Join us in creating the standard for agent-website interaction.

View on GitHub

Want WebMCP built out, or handed over?

This is not a purchase and there is no card field. It puts your address, this product, and whatever you write below in front of a person, and you get a written answer about what finishing it, or handing it over for you to run yourself, would actually take.

A person reads these. Expect a reply in days, not minutes.
Your address is used to answer you. No sequence, no list, no sharing.
Ask to be removed and you are removed.

WebMCP Protocol