# workerd — Cloudflare Workers JavaScript Runtime

> workerd is the open-source JavaScript and WebAssembly runtime that powers Cloudflare Workers. It can be self-hosted to run Workers-compatible code locally or on your own infrastructure with the same V8 isolate-based execution model.

## Install

Save in your project root:

# workerd — Cloudflare Workers JavaScript Runtime

## Quick Use
```bash
# Build from source (requires Bazel and clang)
git clone https://github.com/cloudflare/workerd.git
cd workerd
bazel build //src/workerd/server:workerd
# Or use via wrangler for local dev
npx wrangler dev --experimental-local
```

## Introduction
workerd is the open-source runtime behind Cloudflare Workers, released by Cloudflare in 2022. It uses V8 isolates to run JavaScript and WebAssembly workloads with sub-millisecond cold starts and strong per-request isolation. Unlike Node.js, it implements Web Platform APIs (fetch, Request, Response, crypto) rather than Node-specific APIs.

## What workerd Does
- Executes JavaScript and TypeScript using V8 isolates with Web Platform APIs
- Runs WebAssembly modules alongside JavaScript workers
- Provides per-request isolation without the overhead of separate processes or containers
- Implements the Service Workers API and Cloudflare-specific bindings (KV, R2, D1)
- Serves as the local runtime behind wrangler dev for Workers development

## Architecture Overview
workerd is written in C++ using the Cap'n Proto framework for its configuration and RPC layer. Each incoming request runs in its own V8 isolate, providing memory isolation without process-level overhead. The runtime implements a subset of Web Platform APIs (Fetch, Streams, Web Crypto, URL, TextEncoder) and adds Cloudflare-specific bindings. Configuration is defined in a Cap'n Proto schema file that specifies services, routes, and bindings.

## Self-Hosting & Configuration
- Build from source using Bazel with a C++20 compiler (clang 15+)
- Configuration uses Cap'n Proto schema files to define workers and routing
- Bind workers to specific ports and URL patterns via the config
- Environment variables and secrets are injected through the binding configuration
- Can run standalone or be embedded as a library in other C++ applications

## Key Features
- Sub-millisecond cold starts using V8 isolate reuse
- Per-request memory isolation without container overhead
- Web Platform API compatibility (not Node.js APIs)
- WebAssembly support for running compiled code alongside JS
- Same runtime used in production by Cloudflare Workers globally

## Comparison with Similar Tools
- **Node.js** — general-purpose JS runtime with OS-level APIs; workerd focuses on Web Platform APIs with isolate-level isolation
- **Deno** — secure JS/TS runtime; workerd is more specialized for edge/serverless workloads
- **Bun** — performance-focused JS runtime; workerd prioritizes isolation and cold start speed over raw throughput
- **WinterJS** — Wasmer-based JS runtime; workerd uses V8 and has battle-tested production parity with Cloudflare
- **Fastly Compute** — proprietary edge runtime; workerd is open source and self-hostable

## FAQ
**Q: Can I run workerd locally for development?**
A: Yes. The wrangler CLI uses workerd under the hood for local development via wrangler dev.

**Q: Does workerd support Node.js APIs?**
A: Partially. A Node.js compatibility layer is available for common modules like Buffer and crypto, but workerd primarily targets Web Platform APIs.

**Q: Is workerd suitable for self-hosted production use?**
A: workerd is designed for Cloudflare infrastructure. Self-hosting is possible but requires building from source and managing the configuration yourself. Community tooling around self-hosting is limited.

**Q: What is the difference between workerd and miniflare?**
A: miniflare was a Node.js-based Workers simulator. It has been replaced by workerd as the local runtime, providing true parity with the production environment.

## Sources
- https://github.com/cloudflare/workerd
- https://blog.cloudflare.com/workerd-open-source-workers-runtime/

---
Source: https://tokrepo.com/en/workflows/asset-d7bc2488
Author: AI Open Source