Skip to main content
Middleware wraps request handling — logging, caching, rate limiting, and anything custom. Register with server.use(mw) or via FastMCPOptions.middleware: 
import { FastMCP, LoggingMiddleware, RateLimitingMiddleware } from '@prefecthq/fastmcp-ts/server'

const server = new FastMCP({ name: 'my-server' })
server.use(new LoggingMiddleware())
server.use(new RateLimitingMiddleware(100, 60_000))   // 100 requests per minute
This page is under construction. The outline below sketches what it will cover.
  • Hook levels — setup(server) runs once per server instance (register notification handlers here); onRequest(ctx, next) fires for every request without a more specific hook; per-method hooks (onCallTool, onListTools, onReadResource, onGetPrompt, …) take precedence over onRequest.
  • Built-ins — LoggingMiddleware, CachingMiddleware(ttl, keyFn?), RateLimitingMiddleware(limit, windowMs), SizeLimitingMiddleware(maxBytes), ErrorNormalizationMiddleware, CancellationMiddleware.
  • Error semantics — a thrown McpError from middleware propagates as a protocol error (never converted to a { isError: true } tool result); non-McpError from tool handlers becomes a tool execution error.
  • Writing your own — implementing the Middleware interface, ordering, and composing with auth and transforms.