Cloudflare Experiments
Build at the edge. For real.
106 deployable reference implementations across Cloudflare products - real tools with tests, API docs, and a one-click Deploy button. Not Hello World demos.
- MIT licensed
- 106 experiments
- 17 Cloudflare bindings
- 8 categories
Catalog
Browse by category
Grouped the same way as the docs sidebar. Use docs search with tags like ai, d1, r2, or do to filter by binding.
AI & Machine Learning
22 experiments
Workers AI, embeddings, RAG, image generation, and AI Gateway patterns.
AI Website Summary · AI Website Tag Generator · GitHub Repo Explainer
Web Scraping & Parsing
11 experiments
Fetch, HTMLRewriter, metadata extraction, and structured page APIs.
Metadata Extractor · Website to API · Website to llms.txt
Browser Rendering
10 experiments
Fully rendered DOM via Puppeteer at the edge.
Screenshot API · PDF API · Page Metrics
Network & Monitoring
17 experiments
Reachability, DNS, TLS, latency, change tracking, and CORS debugging.
Is It Down · URL DNS Lookup · Email Auth Checker
Edge Platform
18 experiments
Cache, crypto, auth helpers, feature flags, Realtime, and platform APIs.
Edge Cache · Crypto Hash · WebSocket Echo
Compute & Isolation
4 experiments
Containers, Sandbox SDK, Dynamic Workers, and Workers for Platforms dispatch.
Dynamic Worker Runner · Container Echo · Code Sandbox
Storage & Data
11 experiments
R2, D1, KV, Vectorize, Hyperdrive, Pipelines, and mock API patterns.
R2 Storage · Link Shortener · D1 SQL Playground
Stateful & Async
13 experiments
Durable Objects, Cron, Queues, Workflows, Email, and Analytics Engine.
Durable Counter · DO SQLite Notes · Cron Heartbeat
Workflow
From clone to deployed Worker in four steps
Every experiment follows the same Turborepo layout. Pick one, run it locally, deploy it on its own, then adapt the pattern.
- 01
Pick an experiment
Browse by category - AI, scraping, compute, storage - or search docs by binding name like d1 or r2.
- 02
Run locally
Clone the monorepo, npm install, and npm run dev -- --filter=<name>. Wrangler serves on port 8787.
- 03
Deploy to Workers
Use the Deploy button in each README or wrangler deploy. Configure bindings in the Cloudflare dashboard.
- 04
Adapt the pattern
Fork the code into your project. Each experiment is MIT licensed - use it as a starting point, not a dependency.
git clone https://github.com/shrinathsnayak/cloudflare-experiments.git
cd cloudflare-experiments
npm install
npm run dev -- --filter=ai-website-summary
curl "http://localhost:8787/summary?url=https://example.com"Philosophy
Reference implementations, not toy examples
One focused experiment per product capability, with tests, docs, and a Deploy button.
Product-focused
Each experiment maps to a Cloudflare product or platform capability - AI, storage, networking, security, email, media, and more - not Workers alone. Examples typically deploy as Workers that wire those products at the edge.
Single responsibility
One experiment, one capability. Copy a focused reference instead of untangling a monolith.
Independently deployable
Each folder is self-contained with its own wrangler.json, tests, and one-click Deploy button.
Under 60 seconds
Request paths are designed to complete quickly - ideal for learning, demos, and prototyping.
Documentation
Learn how the monorepo works
Guides for getting started, contributing new experiments, and understanding deployment patterns.
Quick Start
Install, run locally, and deploy your first Worker in minutes.
Philosophy
Why experiments stay small, edge-first, and independently deployable.
Self-Hosted
SaaS-style tools you run on your own Cloudflare account.
Adding Experiments
Scaffold a new Worker with the repo layout, tests, and docs checklist.
Code Standards
TypeScript, Hono routes, validation, and JSON error conventions.
Deployment
Deploy buttons, bindings, secrets, and Wrangler workflows.
Architecture
Turborepo layout and how each experiment stays self-contained.
Ready to build at the edge?
Pick one of 106 experiments, deploy it in minutes, and use the source as a reference for your next Cloudflare product build. Everything is MIT licensed.