Cluster and Nodes

An AEGIS cluster is one or more nodes that cooperate to run workloads on behalf of your tenants. The smallest viable cluster is a single hybrid node on a single machine — perfect for local development or a modest production deployment. As you scale, you add more nodes, give them roles, and let the platform route work across them.

Every node has a role that says what it is allowed to do. A controller node hosts the orchestrator brain — the API surface, the workflow engine, the routing pipeline — but does not run agent containers itself. A worker node is a pure execution host: it runs agent runtimes, talks to the gateway, and reports capacity back upstream. A hybrid node does both, which is the default for single-machine clusters. Workloads are routed to nodes based on free capacity and on capabilities — for example, only nodes that have a particular runtime, GPU, or network reachability advertised will receive jobs that require them.

Beyond the core cluster, edge daemons extend AEGIS onto user-installed machines — laptops, lab boxes, customer hardware. Edge hosts join the cluster through a relay and become first-class execution targets, scoped to their owning tenant.

---

Key ideas

• Cluster — one or more cooperating nodes that present a single AEGIS surface.
• Node roles — controller, worker, or hybrid decide what each machine is allowed to run.
• Capability-aware routing — workloads land on nodes that advertise the runtimes and resources they need.
• Edge daemons — user-installed hosts that extend the cluster onto machines outside the data center.

---

Learn more

• Multi-Node Deployment — running clusters with separate controller and worker nodes.
• Infrastructure — cluster topology, networking, and inter-node trust.
• Pod Architecture — how the orchestrator pod and its sidecars are laid out.
• Edge Mode Overview — extending the cluster onto user-installed edge daemons.