Deploying the self-hosted runtime

The runtime ships as a single container image that reads its license at boot. The same image runs in every environment — what you build in development deploys unchanged. Pick the path that matches how your team already works.

One license, injected at boot

A self-hosted runtime is licensed with an offline token. There is nothing to phone home: the runtime verifies the token locally at startup and unlocks the entitlements it grants. Supply it either as:

• the FASTYOKE_LICENSE environment variable, or
• a file mounted at /etc/fastyoke/license.jwt.

No token — or an expired one past its grace window — boots the free tier rather than failing. A lapsed renewal never takes your app down; it just stops the licensed capabilities from scaling up.

Docker Compose (single node)

The fastest path for a single-node install. One docker compose up against a persistent volume runs the runtime with your license injected; a raw binary under systemd is the equivalent for a bare host. Right for air-gapped and in-firewall deployments where you manage the host directly.

Terraform / OpenTofu (AWS)

An infrastructure-as-code module provisions the instance, a dedicated persistent data volume, and networking on AWS — the same declarative workflow your platform team already uses. It is written to run on OpenTofu as well as Terraform, and takes the license as an input that is delivered to the instance at boot.

Helm (Kubernetes)

A chart deploys the runtime into your existing cluster as a single replica. The license is supplied as a standard Kubernetes secret (bring your own, or let the chart create it), and the database is backed by a persistent volume claim. Optional ingress fronts the service when you want external access.

Verify a deployment

However you deploy, confirm the runtime is up and see which entitlements your license granted:

curl http://<host>:8080/api/v1/health

The license block in the response echoes the current status (active, in grace, or free tier) and the entitlements in effect.

Going further

Every path above is single-node — one runtime against one database volume. For high availability, geo, or disaster recovery inside your own network, see the private-cluster topology on the On-Prem page. For the development workflow that precedes any of these, see FastYoke Runtime.