Workspace for becoming the kagent + Agent Substrate expert at Solo.
📝 Write-up: Thousands of AI Agents on Tens of Pods: kagent Agent Substrate — the full story of this repo: the install, the visualizer, the autoscaler, and every gotcha, with verbatim errors and fixes.
Agent Substrate is a Kubernetes-native runtime for AI agents that breaks the
pod-per-agent model. Idle agents are checkpointed (Zstd snapshots) to object
storage and rehydrated on demand onto a small pool of pre-warmed gVisor worker
pods, so one WorkerPool serves far more agents than it has pods. kagent uses it
two ways: SandboxAgent (a Go declarative agent run as a sandboxed actor) and
AgentHarness (a managed sandbox running a third-party coding agent — OpenClaw
or Hermes — that kagent chats with over ACP).
- LEARNING-PLAN.md — the expert track, phased
- notes/substrate-architecture.md — runtime internals
- notes/kagent-integration.md — SandboxAgent, AgentHarness, ACP
- labs/lab1-kind-substrate.sh — hands-on: substrate + kagent + SandboxAgent on kind
- viz/ — Substrate Scope: real-time actor-lifecycle visualizer (simulated + live cluster modes)
| Source | What it's for |
|---|---|
| https://kagent.dev/docs/kagent/concepts/agent-substrate | Concept page — the elevator pitch + CRD surface |
| https://kagent.dev/docs/kagent/concepts/agent-harness | AgentHarness / ACP / OpenClaw + Hermes backends |
| https://kagent.dev/docs/kagent/examples/agent-substrate | kind walkthrough (basis for lab 1) |
| https://learn.agentsubstrate.dev/ | Internals deep dive: topology, resume-actor flow, ateapi |
| https://github.com/agent-substrate/substrate | Source — docs/architecture.md, docs/glossary.md, docs/api-guide.md |
| https://www.masterthemesh.com/solo/kagent-quickstart-kind/ | Tom O'Rourke's field labs (kagent POC patterns) |