Self-hosting overview
The Slipway control plane is source-available (FSL-1.1-Apache-2.0) and free to self-host on every tier. The runner, CLI and SDKs are Apache-2.0. You get the same product as Slipway Cloud, and your pipelines, secrets, logs and artifacts stay in your infrastructure.
Architecture
Section titled “Architecture” Browser ──HTTP/2──▶ ┌────────────────────────────────────────────────────────────┐ CLI / SDK ─Connect▶ │ server Connect API, auth, webhooks, SPA, live log tail │ SCM webhooks ─────▶ │ scheduler trigger evaluation, DAG to jobs, queue, leases │ │ worker notifications, webhooks, retention, metering │ └───────┬───────────┬───────────┬───────────┬────────────────┘ │ │ │ │ Postgres 17 NATS S3-compatible Valkey (RLS by org) JetStream (logs, (sessions, (queues, artifacts, rate limits) events) caches) runners (shell | docker | kubernetes) ── outbound only ──▶ server- One binary:
slipway server | scheduler | worker | runner | all.allruns server, scheduler and worker in one process for development and small installs. - Stateless components. Server, scheduler and worker scale horizontally. State lives in Postgres, NATS and object storage.
- Runners are separate and dial out. Put them next to the workloads they build and deploy.
Requirements
Section titled “Requirements”| Dependency | Version | Notes |
|---|---|---|
| PostgreSQL | 17 | Primary and optional replicas. The app role is unprivileged so row-level security applies. Managed services (RDS, Cloud SQL, Azure Database) work |
| NATS | 2.11 with JetStream | Clustered (3 nodes) for production |
| Object storage | Any S3 API | AWS S3, MinIO, R2, GCS via interoperability |
| Valkey (or Redis) | 8 | Sessions and rate limits. rediss:// for TLS |
| Compute | Small: 2 vCPU and 4 GB for everything. Production: 3 replicas of server, 2 each of scheduler and worker, sized from the Helm defaults | |
| DNS and TLS | A public HTTPS URL (SLIPWAY_BASE_URL). Required for SCM webhooks and for OIDC federation |
Deployment shapes
Section titled “Deployment shapes”| Shape | When | Guide |
|---|---|---|
| Laptop or single host with Compose | Evaluation, small teams | Compose |
| Kubernetes with the Helm chart | Production | Helm |
| Terraform reference stacks (AWS, GCP) | Production from scratch with managed services | Terraform |
| Binary plus your own dependencies | Bare metal, custom orchestration | Configuration |
Operating it
Section titled “Operating it”- Configuration reference: every
SLIPWAY_*setting. - Upgrades: migrations and rollbacks.
- Backups: what to back up. The master key is critical.
- High availability: replicas, failure modes, scaling.
- Observability: metrics, traces, logs, health.
Licensing in practice
Section titled “Licensing in practice”You can run, modify and use Slipway internally or to deliver your own software without restriction. You cannot offer it to third parties as a competing hosted service. Each release converts to Apache-2.0 two years after publication. Paid tiers on self-hosted installs add support, SLAs and features such as BYOK, change management and SAML/SCIM.