What the agent does
The agent reads the Worker before it adds a line of logging. In a Worker like nodejs-http-server-template in cloudflare/templates, wrangler.jsonc names the Worker, points main at src/index.ts, sets compatibility_date to 2025-10-08 with the nodejs_compat flag, turns observability on and uploads source maps. src/index.ts builds a server with createServer from node:http, answers / with HTML, /api/status with a JSON body of status and timestamp, and anything else with a plain 404, then calls server.listen(8080) and exports httpServerHandler({ port: 8080 }) from cloudflare:node. package.json pins wrangler 4.136.1, typescript 5.9.3 and @types/node 24.10.1, with a check script that runs tsc && wrangler deploy --dry-run and no test script. Nothing logs today, so a request that fails leaves no trace a developer can search for.
Because the Worker uses the Node HTTP server API, the agent wraps the request listener rather than touching each route. It adds src/request-id.ts with one pure function that takes the incoming headers and returns an id: the caller's x-request-id when it is present and matches a conservative pattern of up to 64 URL-safe characters, otherwise a fresh crypto.randomUUID(). The listener in index.ts calls it first, sets x-request-id on the response before any writeHead, records performance.now(), and on the response's finish event writes a single console.log(JSON.stringify({...})) with the id, the cf-ray header Cloudflare adds, the method, the path, the status code and the duration in milliseconds. With observability already enabled in wrangler.jsonc, Workers Logs indexes those JSON fields, so a developer can filter on requestId in the dashboard or wrangler tail and see the one line for a failed call.
The agent adds vitest as a dev dependency with a test script and writes src/request-id.test.ts covering the three cases: a valid incoming id is echoed, an oversized or odd-character id is replaced, and no header yields a UUID. It leaves the three routes and their bodies as they are, so the /api/status response does not change shape for existing callers, and it does not add a logging library, because one console.log of a JSON string is what Workers Logs wants.
Before pushing, the agent runs npm ci, npm run check, which type-checks and performs a dry-run deploy to show the bundle still builds under nodejs_compat, and npx vitest run. It then starts wrangler dev in its sandbox, sends requests to /, /api/status and a missing path with and without an x-request-id header, shows the header on each response and the JSON line wrangler dev prints for each, and quotes those in the pull request.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "Add request id logging to this Cloudflare Worker and open a pull request. Read wrangler.jsonc, package.json, tsconfig.json, worker-configuration.d.ts and src/index.ts first. Add src/request-id.ts with a pure function that returns the incoming x-request-id when it is at most 64 URL-safe characters and otherwise a fresh crypto.randomUUID(). In the request listener, set x-request-id on every response before writeHead and, on the response finish event, write one console.log of a JSON string with the request id, the cf-ray header, method, path, status and duration in milliseconds, so Workers Logs indexes the fields. Do not change the routes or their bodies and do not add a logging library. Add vitest with a test script and tests for the echoed, replaced and generated id cases. Run npm run check and the tests, then start wrangler dev and show the header and the log line for each route with and without an incoming id. Quote the outputs in the pull request.",
"resources": [
{
"url": "https://github.com/your-org/your-app"
}
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
What comes back is a pull request on its own branch that changes src/index.ts to wrap the listener, adds src/request-id.ts and src/request-id.test.ts, and changes package.json and package-lock.json to add vitest and a test script. The description shows a sample log line, explains the header contract, and names the dashboard filter a developer types to find a request by id.
The proof is the output of npm run check and npx vitest run, and the wrangler dev session with the curl responses showing x-request-id on each and the JSON line printed for each, all quoted in the description and kept in the run's record. Whatever workflow the repository runs on pull requests runs here as well. No deploy is claimed; wrangler deploy --dry-run only shows that the bundle builds.
Before merging, a reviewer checks that trusting an incoming x-request-id is wanted, since a client can then choose the value that appears in logs, that the log line holds no header a privacy review would object to, and that the field names match any convention their other services use. Then they deploy and filter Workers Logs on one id from a real request.
Run it on every Worker without request logging
A team with several Workers sends one POST /o/{org}/sessions per repository, with this prompt and the repository in resources. GET /o/{org}/github lists every repository the Rush GitHub App installation can reach, so a script can start a run for each one with a wrangler.jsonc or wrangler.toml and no x-request-id in its source. Each run reads that Worker's own entry point, so a Hono Worker gets a middleware and a plain fetch handler gets a wrapper around the handler, and each opens its own pull request.
Set a webhook URL on the API key and Rush POSTs run.completed or run.failed when a run's turn finishes, with session_url so your server can read the pull request link. Send a fresh UUID as request_id with each request so a retry returns the run already started.
The use case this job follows, with its automation and trigger: Hand off a long task.
Why run it on Rush?
Adding an id and a log line is small, and the questions around it are what take the time: where to set the header so it survives writeHead, when the finish event fires relative to the status being known, what Workers Logs will index and what it will flatten, and whether the bundle still builds under nodejs_compat once node:crypto or the global crypto is in play. On Rush the agent answers those by running wrangler dev in its own sandbox and sending real requests, so the pull request shows the header and the log line rather than describing them.
Every command, response and log line stay in the run's record, the diff is three small files plus the lockfile, and nothing deploys until a person merges. The same request can add the same contract to every Worker a team runs, one sandbox and one pull request each, so request ids line up across services.
Frequently asked questions
Why accept the caller's x-request-id at all?
So a request can be traced across services: a front-end or gateway that already sets the header gets the same id back and in this Worker's logs. The agent bounds it to 64 URL-safe characters so a client cannot inject newlines or a very long string into the log line. If you would rather always generate the id, say so in the prompt and the function shrinks to one line.
Does this need a logging library?
No. Workers Logs, enabled by observability in wrangler.jsonc, parses a console.log of a JSON string into fields you can filter on. A library would add bundle size and another API to learn for a single line per request.
What about the cf-ray header?
Cloudflare attaches a cf-ray value to every request at the edge, and the log line records it beside the request id, so a support ticket that quotes a Ray ID from an error page can be matched to the application's own id. The Worker does not change or return cf-ray.
Will the test run the Worker itself?
The Vitest tests cover the id function, which is where the decisions are. The end-to-end check is the wrangler dev session in the sandbox, with the curl output in the pull request. If you want tests that run the Worker in workerd, ask for @cloudflare/vitest-pool-workers in the prompt and the agent adds that configuration too.
More Cloudflare Workers jobs
The same job on other stacks
- Add Request ID Logging to a CodeIgniter 4 App (CodeIgniter)
- Add Request ID Logging to an Echo App (Echo)
- Add Request ID Logging to a FastAPI App (FastAPI)