TypeScript and Node.js

Backend development in TypeScript and Node.js

Many products need a backend that is quick to build, easy to understand, and simple to operate. We build APIs and services in TypeScript on Node.js 22 and 24, with Express, Zod validation, and typed contracts from OpenAPI. Projects are set up with pnpm, run in Docker, and are tested with Vitest, Supertest, and node:test.

Because the types follow the data from the database to the browser, many errors are caught while the code is being written. That makes the solution safer to change over time, even when several teams work in the same codebase.

Architecture choices in Node.js backends

Node.js offers a lot of freedom, which is exactly why we make some decisions early:

• Contract first. The API is described in OpenAPI before it is implemented. openapi-typescript generates types for the clients, and the specification doubles as documentation.
• Zod at every boundary. Requests, responses from external APIs, environment variables, and webhooks are validated with Zod. Inside the service, the code can then trust the types.
• Thin framework, plain logic. Express handles HTTP, while business logic lives in ordinary functions that can be tested without starting a server.
• The right runtime. Small, stateless services and webhooks fit well on Cloudflare Workers. Services with long-running jobs, heavy dependencies, or a need for the full Node.js platform are packaged in Docker.
• pnpm workspaces. When several services and clients share types, we keep them in a monorepo with pnpm workspaces rather than copying code between projects.

Node.js is a good default for APIs and integrations. When a service has strict requirements for performance, memory use, or unattended operation over long periods, we consider Rust or Go instead.

Automation and tools for development teams

TypeScript and Node.js are well suited to internal tooling. We have built bots that notify the team in Microsoft Teams about deploys and new app versions, using data from Azure DevOps, App Store Connect, and Google Play. We also build MCP servers that give AI agents scoped access to internal systems, and use Playwright and Puppeteer to automate tasks in the browser. The goal is for the team to spend less time on routine work and more time on the product.