The Default Backend / API Stack
First, a money-saving truth: most beginners don't need a separate backend. Next.js API routes (see the web stack) run server code right alongside your app and cover the vast majority of projects. This page is for when you do need a standalone API — a service shared by a mobile app and a web app, or a backend with no web front end at all.
Read the Universal Toolbelt first — these are the backend-specific picks on top of it. And keep the rule above the stack in mind: buy or glue, don't build — a managed database, hosted auth, and a jobs service beat operating that infrastructure yourself.
| Slot | Pick | Why this one | Without it you'd... | Link |
|---|---|---|---|---|
| Language | TypeScript (Node) | Same language as your front end, so you share types and skills across the whole stack. | Context-switch to a second language and re-implement your data types on the server. | TypeScript docs |
| Framework | Hono | Tiny, fast, fully typed web framework that runs anywhere (use Next.js API routes instead when you're full-stack). | Hand-parse requests and responses and lose type safety at your API boundary. | Hono docs |
| Database | Supabase (Postgres) | Managed Postgres with a free tier and a dashboard — same database as the rest of this guide. | Provision, secure, and back up a database server yourself. | Supabase docs |
| ORM | Drizzle | Type-safe SQL in TypeScript with autocomplete, so queries are checked at compile time. | Concatenate SQL strings by hand and open yourself to injection bugs. | Drizzle docs |
| Auth | Clerk / Supabase Auth | Managed auth with a free tier; Supabase Auth is handy when you're already on Supabase. | Build and secure token handling yourself and risk leaking accounts. | Clerk docs |
| Background jobs | Inngest | Run scheduled and queued work (emails, retries, cron) with no separate queue server to operate. | Hand-roll a job queue and cron infrastructure, then debug silent failures. | Inngest docs |
| Hosting | Cloudflare Workers / Railway | Cloudflare Workers for serverless APIs on the edge; Railway when you want a long-running server — both deploy from GitHub. | SSH into a VPS and configure the runtime, process manager, and TLS by hand. | Cloudflare Workers docs · Railway docs |
Don't bother yet
These show up in "real backend" tutorials and make you feel like you're missing something. You're not — skip them until scale actually forces the issue.
- Microservices — many small services to coordinate; one service is correct until you have a team and real scale pain.
- Message queues (Kafka, RabbitMQ) — heavy infrastructure for decoupling; Inngest covers background work without running a broker.
- Kubernetes — fleet orchestration you don't need when a platform runs your app for you.
- gRPC — a binary protocol for service-to-service calls at scale; plain HTTP/JSON is simpler and enough.
- Self-hosted databases — operating your own Postgres is a job; let Supabase do it on a free tier.