Koyeb
Serverless compute platform that deploys apps in secure VMs on bare metal servers, with support for Docker, Git, and pre-built containers across multiple cloud regions.
Best for small always-on APIs, background workers, and microservices that need a free instance without cold starts or spin-down timers.
Use Cases
Free Tier
1 free instance: 512MB RAM, 0.1 vCPU, 2GB SSD, 100GB bandwidth/month, no credit card required
How to Maximize the Free Tier
Koyeb's free instance is always-on (no cold starts or sleep timers), making it ideal for small APIs and background workers that need to stay responsive. Deploy via GitHub for automatic CI/CD. The 0.1 vCPU is shared and low-power, so avoid CPU-intensive workloads — use it for lightweight Node.js/Python APIs, static site serving, or webhook receivers. The 100GB monthly bandwidth is generous. Combine with the free Postgres tier (5 hours/month compute, 1GB storage) for a full-stack setup. One free instance per organization, so team accounts share the limit.
Getting Started
Sign up at koyeb.com (GitHub, GitLab, or email) → connect a Git repository or choose a Docker template → select a region (US, EU, or Asia) → configure environment variables and build settings → deploy — Koyeb builds and runs your app automatically → add a custom domain or use the provided .koyeb.app subdomain.
Pros
- Always-on free tier: Free instances don't spin down or sleep — your API stays responsive 24/7 without cron-based wake-up hacks
- Multi-runtime: Supports Docker containers, Node.js, Python, Go, Rust, and any language with a Dockerfile — more flexible than single-runtime platforms
- Global regions: Deploy to US, EU, or Asia regions on the free tier with automatic HTTPS and private networking between services
Cons
- Low CPU: Free instance offers only 0.1 vCPU shared — not suitable for CPU-bound workloads or anything beyond lightweight APIs
- Single instance limit: Only one free instance per organization — team projects or multiple services require paid plans
- Postgres expiration: Free Postgres database expires after 90 days and must be recreated — not suitable for persistent data without a paid plan