I turned my Hono API into an MCP server by changing one import

Description and use

Convert a Hono API into an MCP server by changing one import, enabling AI agents to call endpoints without duplicating schemas.

Problem

Expose an existing Hono API to AI agents without duplicating endpoint definitions.

Outcome

The bridge turns a Hono app into an MCP server with a single import change, eliminating duplicate tool definitions and preserving existing validation and middleware.

Workflow

• Import mountMcp from hono-mcp-bridge and replace zValidator import.
• Call mountMcp(app) to register every route as an MCP tool.
• The bridge walks app.routes, extracts schemas via z.toJSONSchema, and exposes them as tools.
• Agents invoke tools by sending JSON RPC requests to /mcp; the bridge forwards arguments as a real HTTP request to the original handler.
• Validation, auth, logging, and rate limits are preserved.
• Optional configuration (include, exclude, methods, describeTool, allowedOrigins) limits exposed tools and origins.

Evidence

Method, path, parameters, types, even descriptions: everything an agent needs is already there.

Source

mountMcp(app) // POST /mcp: every route is now a tool

Source

A tool call is a real request. Arguments become a Request passed to app.fetch()

Source

Agents act with the credentials of whoever connected them. Use include, exclude or methods: ['GET'] to limit what is exposed

Source