Skip to main content
WebMCP gives websites a browser-native way to publish structured tools through document.modelContext. The tools run inside the page’s origin, authenticated session, application state, and visible browser context. That location is the point of the design. Agents get an explicit capability contract from the site instead of inferring intent from screenshots, DOM shape, or brittle automation scripts. The W3C WebMCP specification is authoritative for the standard surface. MCP-B supplies implementation and integration tooling, including the strict polyfill, React bindings, browser transports, iframe bridges, local relays, and extension tooling. MCP-B capabilities beyond the standard tool surface are extensions. For implementation, follow Build your first tool and Choose a runtime.