Koyeb
High-performance Infrastructure for APIs, Inference, and Databases·koyeb.com
Koyeb is a global, high-performance serverless platform for deploying and scaling AI applications, APIs, and databases across CPUs, GPUs, and accelerators. It abstracts server and infrastructure management, offering autoscaling, scale-to-zero, zero-downtime deployments, and global deployment in over 50 locations. Users can deploy via Git push, Docker containers, CLI, or API, and run inference, sandboxes for AI agents, and production workloads without managing servers. It also offers a fully managed Serverless Postgres database.
What it's for
Features 17
- Autoscaling with scale-to-zero
- CLI, API, Terraform, and Pulumi support
- Deploy from Git push, Docker containers, or pre-built images
- Global deployment across 50+ locations
- Instant API endpoint provisioning
- Koyeb Sandboxes for isolated AI agent code execution
- Manual scaling of services
- Multi-factor authentication and passkeys for accounts
- Native HTTP/2, WebSocket, and gRPC support
- One-click app deployments (AI models, full-stack apps, databases)
- Real-time logs and direct instance access
- Serverless GPU, CPU, and accelerator compute (AMD, Intel, Nvidia)
- Serverless Postgres with pgvector support
- Service discovery and service mesh
- Sub-200ms cold starts
- Ultra-fast NVMe storage
- Zero-downtime deployments with automatic health checks
Pricing
Pro and Scale plans include a base monthly fee plus metered compute usage; Enterprise is quote-based, described as 'starting at $1000/mo' plus compute. Compute is billed per second across CPU, GPU, and accelerator instance types with rates from $0.50/hr to $24.00/hr. Serverless Postgres is billed separately by instance size and storage.
Base subscription fee; additional compute usage billed separately (+compute).
$10 included compute, 10 users, 100 services, 5 concurrent builds
Base subscription fee; additional compute usage billed separately (+compute).
$100 included compute, 50 users, 1000 services, AWS Regions, Slack cross-connect, 99.9% uptime SLA
Custom quote-based pricing; page states plans 'starting at $1000/mo' plus compute usage.
Custom, starting at $1000/mo +compute; unlimited users
Long-term reservations (saving plans) get up to 50% off the on-demand price.