Skip to main content
This section covers the why behind WebMCP. It is designed for reading when you step back from the code and want to understand the bigger picture. You will not find step-by-step instructions here. For that, see Tutorials and How-To Guides. For precise API descriptions, see Packages.

Concepts

What Is WebMCP?

The proposed API, why it exists, and how it relates to AI agents in the browser.

WebMCP vs MCP

How WebMCP relates to the server-side Model Context Protocol.

Native vs Polyfill vs Global

Three runtime strategies and when each one applies.

Strict Core vs MCP-B Extensions

Where the specification ends and the MCP-B extension layer begins.

Architecture

Runtime Layering

How native APIs, the polyfill, and BrowserMcpServer stack on top of each other.

Transports and Bridges

How tools cross tab, iframe, extension, and desktop boundaries.

Tool Lifecycle and Context Replacement

Registration, discovery, context replacement, and why ordering matters.

Design

Tool Design

Principles for naming, schemas, composability, and failure handling.

Security and Human-in-the-Loop

The safety model, user presence, and browser-mediated trust.

Spec Status and Limitations

Where the specification stands and what is still changing.