WebMCP in brief
WebMCP (Web Model Context Protocol) is a planned W3C web standard from the Web Machine Learning Community Group. It lets developers run their web apps as MCP servers with tools executed client-side in JavaScript. AI agents, browser assistants, and assistive technologies can call those tools without a separate backend. WebMCP brings the Model Context Protocol directly to the web.
What is the Model Context Protocol (MCP)?
The Model Context Protocol (MCP) is a specification that connects AI systems like ChatGPT, Claude, or Gemini to external tools and data sources. Think of a USB-C port for AI: a standard way for large language models (LLMs) to call tools, fetch data, or trigger actions. MCP servers usually run as separate processes—WebMCP moves that logic into the browser.
- Tools:Functions AI agents can invoke—e.g. a calendar tool for booking or a product search tool.
- Prompts:Predefined templates for standardised LLM interactions.
- Resources:Data and content exposed via URI as context for LLMs.
How does WebMCP work?
WebMCP extends the browser API: via navigator.modelContext, web apps can register tools. Each tool has a name, a natural-language description, a JSON schema for inputs, and an implementation. The description helps agents know when and how to use the tool.
- provideContext and registerTool:Tools are registered with the browser—either in bulk via
provideContext()or individually viaregisterTool(). - Client-side execution:Tools run in the page JavaScript context—no extra server, no extra API. You reuse existing application logic.
- User interaction:Tools can involve users via
requestUserInteraction()—e.g. confirmations or inputs—while preserving shared context and user control.
Where WebMCP fits
WebMCP suits sites that want to expose capabilities to AI agents or browser assistants without extra infrastructure. E-commerce can expose product search and advice; booking sites can expose scheduling; editorial systems can expose content workflows.
- AI agents (Claude, ChatGPT, Gemini): Users connect MCP clients such as Claude Desktop to WebMCP-enabled pages—the agent can then call the site tools directly.
- Assistive technologies: Tools can be used by accessibility assistants—for simplified navigation or adapted interactions.
- Browser assistants: Built-in or extension-based agents can invoke site features directly.
Status and standardisation
WebMCP is a draft Community Group report from the Web Machine Learning Community Group—not yet a formal W3C standard. Contributors include Microsoft and Google. The spec targets collaborative workflows: users and agents share the same web UI, context, and control.
- Implementation: Developers add a script and register tools via the JavaScript API. MCP clients connect to the site through their settings.
- Declarative WebMCP: A declarative variant is planned—forms and HTML elements become tools without explicit JavaScript. Still in development.
WebMCP—summary
WebMCP adds a standard interface for AI agents and assistive tech. Sites become MCP servers with client-side tools LLMs can call directly—reducing backend work and enabling new AI-driven scenarios.
IVIS MEDIA builds web applications and integrates modern web standards—from frontend architecture and APIs to AI integration. More about web development. Use our free AI readiness check to see whether your site is ready for AI systems.
Frequently asked questions about WebMCP
What is WebMCP?
WebMCP (Web Model Context Protocol) is a planned W3C-oriented standard that lets websites expose AI-callable tools. Developers register JavaScript functions as tools—AI agents, browser assistants, and assistive technologies can invoke them directly, so the site behaves like an MCP server running in the browser.
What is the difference between MCP and WebMCP?
MCP is the general specification—servers often run as separate processes or services. WebMCP brings MCP to the web: tools are implemented client-side in JavaScript; no dedicated server is required. The website itself acts as the MCP server.
What is WebMCP good for?
WebMCP fits web apps that want to expose features to AI agents or assistive tech—e.g. e-commerce search, booking, editorial or research tools. Users can connect MCP clients like Claude Desktop to WebMCP-enabled pages and use those features directly.
How do you implement WebMCP?
Via the navigator.modelContext API, tools are registered with name, description, and JSON schema—using provideContext() or registerTool(). The handler runs in the page JavaScript context. Add a script, register tools, then connect MCP clients to the site.
Is WebMCP an official standard?
WebMCP is a Draft Community Group report from the Web Machine Learning Community Group—not yet a formal W3C standard. It is actively developed with participation from Microsoft and Google.
How secure is WebMCP?
The WebMCP API is defined as a secure context—it runs over HTTPS. Tools execute in the site context; user interactions can be requested via requestUserInteraction(). Secure implementation and user control are part of the security and privacy model.