Bridge package that turns any oRPC backend into browser-native AI tools via the WebMCP spec — zero per-endpoint config. Next.js, TypeScript, Zod.
WebMCP is a W3C Community Group spec that moves AI tool integration into the browser. This demo exposes a Next.js + oRPC backend's API routes as WebMCP tools, so AI agents call them natively inside Chrome.
The orpc-webmcp bridge package registers backend procedures as browser-accessible tools automatically — no per-endpoint config.
navigator.modelContext API in Chrome Canary (146+)Write type-safe procedures with oRPC and Zod. Standard API work, nothing WebMCP-specific.
The orpc-webmcp bridge discovers every procedure and registers it as a WebMCP tool via navigator.modelContext.
Any in-browser AI agent — like Chrome's built-in AI — discovers and calls your API endpoints as tools.
Calls validate against Zod schemas, execute server-side via oRPC, and return results to the agent. Fully type-safe.
WebMCP runs today in Chrome Canary on the built-in Gemini Nano model. Five steps:
chrome://flags/#web-mcp to Enabledgit clone https://github.com/Mark-Life/webMCP-example && cd webMCP-example && bun install && bun devhttp://localhost:3000 in Canary — the app registers its tools via navigator.modelContextNo polyfills. Native navigator.modelContext API, Chrome 146+.
The demo task manager exposes five tools:
apps/web/ — Next.js app: UI, server actions, oRPC route handlerpackages/orpc-webmcp/ — the bridge package: core logic and React hooksorpc-webmcp is the core: a generic adapter that:
navigator.modelContextAny oRPC app becomes WebMCP-compatible with a single hook.

Open source. Both a reference implementation of the WebMCP spec and a starting point for browser-native AI integrations.