HostingCAN I VIBECODE
Cloudflare Workers?
BUILD FIRST
- A non-sensitive or read-only version of the core experience
- A manual review step wherever correctness or trust matters
- A narrow companion tool that removes a real piece of work
LEAVE FOR LATER
- Regulated decisions, payments, or irreversible actions
- Claims of correctness, compliance, or uptime that the MVP cannot support
- Sensitive integrations until security, support, and auditability are ready
RESEARCH NOTES
What this tool actually does.
Cloudflare Workers is a serverless compute platform that runs JavaScript, TypeScript, Python, Rust, and WebAssembly at the edge of Cloudflare's global network (330+ cities). Unlike traditional hosting, there are no servers, containers, or wall-clock charges: code runs in V8 isolates, lightweight sandboxes that start ~100x faster than containerized Node processes, eliminating cold starts. Deployment uses the wrangler CLI (wrangler dev locally, wrangler deploy to production), with instant rollbacks and gradual rollouts. A Worker exports ES-module handlers such as fetch(), cron triggers, and queue consumers; every request invokes the handler at the nearest data center. Native bindings provide KV (low-latency key-value), R2 (zero-egress object storage), D1 (SQLite), Queues, durable Workflows, and Workers AI (serverless GPU inference for LLMs and image generation). Billing is CPU-time based.
RESEARCHED BUILD PROMPT
Give this to your coding agent.
Build a minimal, self-contained replacement for the core developer workflow of Cloudflare Workers (the serverless edge-compute product), without copying Cloudflare's brand, logos, documentation UI, or code. Target workflow: a developer writes a Worker (an ES-module file exporting a fetch(request) handler), runs it locally with hot reload, then deploys it to a simulated edge and invokes it by URL.
User-facing requirements: a CLI (name it something original, e.g. edgework) with dev, deploy, list, delete, and logs commands. `dev` watches the worker file, serves it on localhost, injects a minimal Request/Response Web API surface, reloads on save, and prints per-request logs. `deploy` publishes to a local registry, assigns a stable URL like http://<name>.edgework.test, and supports instant rollback to a previous version. Show a fake CPU-time billing meter after each request. Support a TOML config file declaring routes and cron triggers. Add a tiny per-worker key-value store available to handlers.
Technical approach: implement the sandbox with Node worker_threads or a vm context exposing fetch, Request, Response, Headers, and crypto from Node globals — no child process per request. Persist versions, routes, and KV to JSON files under a project .edgework directory; no external database. Route requests in-process via hostname mapping; run cron triggers with node-cron or a simple setInterval scheduler.
Deliberately out of scope: a real CDN/edge network, R2/D1/Queues/Workers AI bindings, a dashboard UI, user auth or multi-tenant platform, and a WASM toolchain.
Deliver a README covering setup, all commands, an example worker, how the sandbox isolates user code, and important limitations (no true 330-city deployment, approximate billing, single-machine storage).
Do not copy the original brand, proprietary assets, or UI; create an original visual system.