After this lesson you'll be able to name all 16 products on Cloudflare's Developer Platform, say what each one is for in a sentence, and pick the right one from a one-line decision guide — without needing to know how any of them work internally yet.
Cloudflare's Developer Platform is not one product — it's 16, grouped into four categories: Compute, Media, AI, and Storage & Databases. They all run on the same global network, and most real applications combine several of them (a Worker calling D1 and R2 is a completely normal stack, not an edge case). This lesson is the map. It intentionally stays shallow — one lesson per product (or more, where a product is complex) will go deep on mechanics, pricing, and worked examples later in this course.
Compute is where your application's logic runs. Six products, ranging from "run this function" to "let your customers run their own code on your platform."
Workers is Cloudflare's serverless execution environment: you deploy JavaScript/TypeScript (or WASM) and it runs on Cloudflare's network close to the requesting user, not in one region. Reach for it as the default backend for HTTP APIs, full-stack apps, and anything else you'd otherwise put on a traditional serverless function — it's the foundation almost every other product on this list is either built on or invoked from.
Durable Objects are single-instance, stateful "actors" — each one has its own persistent storage and processes requests one at a time, so you get strong consistency without a separate coordination layer. Reach for them when you need real-time, stateful coordination between multiple clients: chat rooms, collaborative editing, multiplayer game state, WebSocket hubs, rate limiters.
Observability is Cloudflare's built-in logging, tracing, and analytics layer for Workers — logs, real-time tail, and metrics without wiring up a third-party APM from day one. Reach for it as soon as you have anything in production, since it's how you'll debug a Worker you can't attach a local debugger to.
Workers for Platforms lets you run other people's code — your customers' scripts — safely and cheaply inside your own product, using isolate-level sandboxing rather than one container per tenant. Reach for it when you're building a platform that needs to be programmable by its own users (think: a SaaS product offering a "custom logic" or "plugins" feature).
Workflows adds durable, multi-step execution on top of Workers: each step's state is checkpointed, so long-running or multi-stage processes automatically resume and retry after failures instead of restarting from scratch. Reach for it for anything that looks like a pipeline — order processing, multi-stage AI agent tasks, data ingestion — where a single request/response Worker isn't a natural fit.
Pages is Cloudflare's platform for full-stack web apps and static sites built with popular frameworks (Next.js, Astro, Remix, etc.), with git-based deploys and preview URLs out of the box. Reach for it when you want framework-native, zero-config hosting for a frontend-heavy app — it deploys onto the same underlying network as Workers.
Three products for handling images, video, and real-time audio/video at scale, so you don't build your own transcoding or delivery pipeline.
Images is a managed pipeline for resizing, optimizing, storing, and serving images on demand via URL parameters, instead of you pre-generating every size and format. Reach for it whenever your app needs responsive image delivery — product photos, avatars, user uploads — without hand-rolling a resize-and-cache system.
Stream ingests, encodes, stores, and plays back both live and on-demand video, handling adaptive bitrate and a built-in player so you don't manage encoding infrastructure. Reach for it for anything video: course platforms, user-generated video, live streaming features.
Realtime provides the infrastructure (based on WebRTC, via Cloudflare's Calls/Realtime SDKs) for low-latency audio and video communication between users. Reach for it when you're building video calls, voice chat, or other immersive real-time audio/video experiences directly into your app.
Three products for running, managing, and searching around AI models, whether the model runs on Cloudflare or elsewhere.
AI Gateway sits in front of your calls to any LLM provider (OpenAI, Anthropic, Workers AI, etc.) and adds caching, rate limiting, retries, logging, and cost analytics without changing your application code. Reach for it as soon as you're calling an LLM API in production and want visibility into cost and usage, or a cache to cut repeat-prompt spend.
Vectorize is a globally distributed vector database for storing and querying embeddings, built to sit close to your users without you managing sharding or scaling. Reach for it any time you're building semantic search or retrieval-augmented generation (RAG) and need to store embeddings alongside metadata.
Workers AI runs inference for common open models (LLMs, image generation, speech-to-text, embeddings, and more) directly on Cloudflare's network, billed per-request rather than per-GPU-hour. Reach for it when you want AI inference without provisioning or managing your own model-serving infrastructure, or when you want inference to run close to the user for latency.
Five products, and the ones most likely to get confused with each other, because "where do I put my data" has more than one right answer depending on the shape of that data and how it's accessed.
D1 is Cloudflare's serverless SQL database (SQLite-based), with point-in-time recovery, for data that genuinely needs relations, joins, and transactions. Reach for it for the kind of data you'd normally put in Postgres or MySQL on a small-to-medium app: users, orders, posts.
Hyperdrive isn't a database — it's a connection-pooling and caching proxy that makes an existing regional database (Postgres, MySQL) usable from Workers without eating a slow round-trip and a fresh TCP/TLS handshake on every request. Reach for it when you already have a database outside Cloudflare and want to call it from a Worker without a rewrite.
R2 is S3-compatible object storage for files and blobs, with the standout feature being zero egress fees. Reach for it for anything file-shaped: user uploads, backups, static assets, datasets — especially if you'll be reading that data a lot, since egress is usually where object storage bills get expensive elsewhere.
Workers KV is a key-value store replicated to Cloudflare's edge locations for fast reads everywhere, at the cost of being eventually consistent — a write isn't guaranteed to be visible everywhere instantly. Reach for it for read-heavy, rarely-changing data: config, feature flags, session lookups, cached API responses.
Queues lets Workers push messages onto a queue and process them asynchronously with a consumer Worker, decoupling "accept the request" from "do the slow work." Reach for it whenever you need to smooth out load, retry failed work, or fan out a task without making the original caller wait for it to finish.
Each of the 16 products above gets its own lesson (or, for the more complex ones — Workers, Durable Objects, D1 — more than one), covering mechanics, pricing, use cases, and a worked code example in depth. This overview is the only lesson that treats the platform in one wide, shallow pass; everything after it goes narrow and deep, one product at a time, starting with the product almost everything else depends on: Workers.
Cloudflare Developer Platform — Products is the canonical page defining this 16-product, 4-category lineup, and is the page to re-check if you suspect the lineup has changed since this lesson was written (2026-07-03).
R2 vs. Workers KV: R2 is object storage for files/blobs (think S3, with zero egress fees); KV is a key-value store for small, read-heavy, eventually-consistent data like config or flags. They solve different shapes of data, not the same problem at different scales.
D1 vs. Hyperdrive: D1 is a database (serverless SQLite) that Cloudflare hosts for you. Hyperdrive is not a database — it's a proxy that makes a database you already run elsewhere (Postgres/MySQL) fast to reach from a Worker.