Your Website Becomes the MCP Server. No Separate Deployment. No Scraping.
WebMCP is the W3C browser standard, co-authored by Google and Microsoft, that exposes structured tools from your website directly to AI agents via the navigator.modelContext API. eComerfy MCP is South Africa's first agency implementing it.
What Is WebMCP?
The Web Standard That Lets AI Agents Act on Your Website
Right now, AI agents trying to interact with websites must crawl, scrape, and reverse-engineer how everything works. To book a flight, an agent identifies input fields, guesses the data format, and hopes nothing breaks. This is fragile, unreliable, and about to change.
WebMCP is the common language. A W3C Community Group Draft, currently in early preview in Chrome 146, WebMCP uses the navigator.modelContext browser API to let websites register structured tools, search, book, price-check, enquire, that AI agents can discover and invoke without guessing.
Unlike server-side MCP which requires separate infrastructure, your existing website becomes the MCP server. Tools run in the browser tab. No cloud deployment. No separate server. Just structured tools your site registers, and AI agents execute.
W3C Community Group Draft
Open Standard, Not Proprietary
Co-authored by engineers from Google and Microsoft. When two major browser and AI vendors co-author the spec, the direction is set.
Chrome 146 Beta
Live in the Browser Today
Enable the WebMCP flag in Chrome 146, install the MCP Tool Inspector Extension, and test your tools with real AI agent interactions, including Gemini API support.
MCP Ecosystem: 2026
97M Monthly SDK Downloads
10,000+ active MCP servers. First-class support in Claude, ChatGPT, Gemini, Copilot, Cursor, and VS Code. MCP donated to the Linux Foundation's Agentic AI Foundation (AAIF), cementing vendor-neutral governance.
Technology Comparison
RAG vs MCP vs WebMCP
Understanding where WebMCP fits in the AI infrastructure stack, and why it represents a fundamentally different approach to agent-site interaction.
| Feature | RAG | MCP (Server) | WebMCP |
|---|---|---|---|
| Where it runs | AI backend infrastructure | Separate MCP server | In the browser tab |
| Access method | Vector similarity search | JSON-RPC API call | navigator.modelContext API |
| Setup complexity | High | Medium–High | Low (browser-native) |
| Agent interaction | Read-only retrieval | Tools + actions | Structured browser tools |
| Deployment requirement | AI infrastructure | Cloud or local server | Your existing website |
| Schema.org pairing | Recommended | Optional | Essential, describes what WebMCP tools interact with |
Our Process
How We Implement WebMCP
A structured, spec-aligned process from audit to live tool deployment, built around the W3C draft and Chrome 146 preview.
Step 01
Audit Your Key Actions
We identify every high-value action users take on your site, product search, booking, price lookup, enquiry form, and assess which should be exposed as WebMCP tools.
Step 02
Design Tool Definitions
We map each action to a declarative or imperative WebMCP tool schema, defining inputs, outputs, and the natural language descriptions AI agents use to discover and invoke them.
Step 03
Implement & Register
We implement tool registration via the navigator.modelContext API and wire it to your backend data systems, inventory, booking engine, CRM, or catalogue.
Step 04
Test with Chrome Inspector
We test every tool using Chrome 146's MCP Tool Inspector Extension, verifying discovery, parameter handling, response structure, and edge cases.
Step 05
Layer Schema.org Markup
We pair every WebMCP tool with the corresponding Schema.org vocabulary, so AI agents understand both what your content is and how to interact with it.
Step 06
Monitor & Evolve
As the W3C spec and Chrome support matures, we update your implementation, keeping you ahead of every competitor still waiting for the standard to stabilise.
WebMCP Services
What We Deliver
From readiness audits to full implementation, everything your site needs to be discovered and activated by AI agents.
WebMCP Readiness Audit
A thorough audit of your current website against the WebMCP specification, identifying which of your key actions should become WebMCP tools and what technical work is needed.
WebMCP Tool Design
Map your business's high-value actions, product search, booking, price-check, enquiry, to declarative or imperative WebMCP tool definitions optimised for AI agent interaction.
WebMCP Implementation
Register structured WebMCP tools via the navigator.modelContext API. We implement, test with Chrome 146's MCP Tool Inspector, and validate tool responses against the W3C draft spec.
Schema.org + WebMCP Stack
Schema describes what your content is. WebMCP defines how AI agents interact with it. We implement both layers, full JSON-LD schema markup paired with WebMCP tool definitions.
Server-Side MCP Integration
For complex workflows requiring backend data access, we build the supporting MCP server architecture, product catalogues, booking systems, CRMs, that WebMCP tools call into.
Security & Access Control
WebMCP tools that expose sensitive actions need OAuth 2.1 protection and rate limiting. We implement the full security layer aligned to the latest MCP spec (2025-06-18 release).
AI Agent Discoverability
Ensure AI agents can find and trust your WebMCP tools, server cards, agent discovery files, .well-known endpoints, and llms.txt, so your site is visible to every major AI platform.
Ongoing Evolution & Monitoring
The WebMCP spec is live and evolving. We monitor W3C updates, Chrome releases, and MCP protocol changes, keeping your implementation current as the standard matures.
Why Act Now
This Is the Mobile-First Moment, for AI
In the same way mobile-first design redefined how websites were built, agent-ready design will define the next generation of the web. WebMCP is the spec that makes agent-readiness concrete and standardised.
With Google and Microsoft co-authoring the spec, Chrome 146 shipping it behind a flag, and the MCP ecosystem at 97 million monthly SDK downloads, the direction is not speculative. The window to be an early mover is open now.
Businesses that register WebMCP tools early will be discoverable and actionable in AI-agent-assisted research before competitors even know the standard exists. That's compounding advantage, the same dynamic that rewarded early SEO adopters in 2003.
Monthly MCP SDK downloads across Python + TypeScript
Active MCP servers globally, tools, APIs, and data sources
WebMCP live behind flag, Google + Microsoft co-authoring spec
MCP donated to AAIF, Anthropic, OpenAI, Block co-founding
The Complete AI-Readiness Stack
Schema.org + WebMCP
Schema describes what your content is. WebMCP defines how AI agents interact with it. Together, they form the complete AI-readiness stack every forward-thinking website needs.
The Vocabulary
Schema.org Markup
JSON-LD structured data tells AI systems what your content means, product, service, organisation, event, FAQ. It's the semantic layer that makes your content machine-readable.
- Product schema
- Organisation schema
- Service schema
- FAQ schema
- LocalBusiness schema
The Actions
WebMCP Tools
navigator.modelContext tool definitions tell AI agents what they can do on your site, search products, check availability, request a quote, book a service, get pricing.
- Search tool
- Booking tool
- Price-check tool
- Enquiry tool
- Availability tool
The Signpost
Agent Discovery Files
llms.txt, ai.txt, .well-known/mcp/server-card.json, and agent-card.json ensure AI systems crawling the web know your tools exist before they even load your page.
- llms.txt
- ai.txt
- server-card.json
- agent-card.json
- sitemap.xml signals
FAQ
WebMCP Questions Answered
Get Started
Ready to make your website AI-agent ready with WebMCP?
We'll audit your site against the W3C WebMCP spec, design your tool definitions, and implement navigator.modelContext, so AI agents can discover and act on your business before your competitors know the standard exists.