Cloudflare

1. Prerequisites

Wrangler installed v2.2 or newer. Check your version:

wrangler version

2. Install

npm i @middleware.io/agent-apm-worker

import { init, track } from '@middleware.io/agent-apm-worker';

init({
  serviceName: "{APM-SERVICE-NAME}",
  accountKey: "<MW_API_KEY>",
  target: "https://{ACCOUNT-UID}.middleware.io",
  consoleLogEnabled: false, // set true to see instrumented logs locally
});

3. Capture application data

• Trace the incoming request Use track(request, ctx) and return via sdk.sendResponse(response):

  export default {
     async fetch(request: Request, env: Env, ctx: ExecutionContext): Promise<Response> {
        const sdk = track(request, ctx);
  
        // your app logic
        const response = new Response("ok", { status: 200 });
  
        return sdk.sendResponse(response);
     },
  };

• Trace upstream calls (HTTP client propagation) If your Worker calls an external API, use sdk.fetch(...) to propagate context automatically:

  const sdk = track(request, ctx);
  const url = new URL(request.url);
  const upstream = await sdk.fetch(`https://httpbin.org${url.pathname}`);
  return sdk.sendResponse(upstream);

4. Custom logs (optional)

Emit structured logs correlated with the current trace:

// sdk.logger.SEVERITY(message, attributes?)
sdk.logger.error("error test");
sdk.logger.error("error with attributes", { "log.file.name": "error.log" });
sdk.logger.info("info test");
sdk.logger.debug("debug test");
sdk.logger.warn("warn test");

5. Configure secrets & env (Wrangler)

Use Wrangler vars/secrets to avoid hardcoding keys:

• wrangler.toml

  name = "my-worker"
     main = "src/index.ts"
     compatibility_date = "2024-12-01"
  
     [vars]
     SERVICE_NAME = "orders-worker"
     MW_TARGET = "https://<ACCOUNT-UID>.middleware.io"
  
     # Per-environment overrides
     [env.staging.vars]
     SERVICE_NAME = "orders-worker-staging"
  
     [env.production.vars]
     SERVICE_NAME = "orders-worker"

• Set your API key as a secret:

  wrangler secret put MW_API_KEY
     # optionally per-env:
     wrangler secret put MW_API_KEY --env production

• Then read from env in your Worker:

  init({
     serviceName: env.SERVICE_NAME,
     accountKey: env.MW_API_KEY, // bound secret
     target: env.MW_TARGET,
     consoleLogEnabled: false,
  });

6. View your data

After deployment, give it 3–5 minutes, then open Middleware to see:

• Traces (incoming Worker requests, plus any upstream calls done via sdk.fetch)
• Logs (from sdk.logger.*)
• Profiling is not yet available for Cloudflare Workers in Middleware.

7. Environment variables

(Names/usage align with the Middleware Worker agent options.) (docs.middleware.io, npm)

8. Troubleshooting

• No traces/logs:
  • Confirm init(...) is called and track(request, ctx) wraps every request.
  • Check that accountKey/target are set (or bound via env).
• Upstream service not showing as a child span:
  • Use sdk.fetch(...) instead of fetch(...) to propagate context.
• Local dev secrets:
  • Use wrangler secret put for real secrets and .dev.vars for local-only values.
• Multiple environments:
  • Define vars per environment in wrangler.toml ([env.staging], [env.production]).