Early, greenfield development

Tritolem

A Kubernetes-native, modular platform for providers to run standardized managed services across many existing Kubernetes clusters. Customers order and manage services through one API and CLI, without handling Helm, clusters, or operations themselves.

How it works

A provider defines a service offering once and rolls it out as many governed instances for different customers and projects. A central control plane coordinates policy, placement, and lifecycle; it keeps its state as Kubernetes custom resources and exposes a REST API and CLI. A pull-based agent on each workload cluster carries out the local desired state on its own schedule, so already-running services keep working even when the agent cannot reach the control plane.

Core concepts

Design — not yet implemented

Customers & projects

Providers manage customers and projects, and control which teams may order services within which projects.

Service extensions & offerings

A service extension defines what a service can technically do. Providers turn that into concrete, customer-facing offerings: approved releases, sizes, and options.

Service instances

Each ordered instance is isolated in its own namespace, with its own placement, endpoint, status, and lifecycle.

Clusters & agents

A pull-based agent runs on every connected workload cluster, reconciling the desired state locally and keeping already-running services available even if the control plane is unreachable.

Status

Tritolem is in early, greenfield development. Nothing described on this site is released or generally available yet; the architecture and product scope are still being defined and refined. The documentation tracks what is planned versus implemented as the project progresses.