Roadmap: From understanding to replication

Goal: turn “I can read the code” into “I can rebuild the key skeleton”.

Recommended order

1. Terminology
2. Gateway control plane
3. Agent execution
4. Session concurrency
5. Tool policy + exec approvals
6. Lifecycle + ops

Advanced modules (pick what you need)

For “I want to replicate this architecture” readers:

1. The 5-piece kit
2. Context engineering
3. Model fallback + robustness
4. Memory system
5. Hooks + plugins governance
6. Subagents + skills snapshots

Use one observable baseline to validate each chapter:

• Start: openclaw gateway
• Observe: openclaw logs --follow
• Interact: openclaw dashboard

Hard checks

1. Explain why chat.send ACKs fast and results stream via events (see Control UI).
2. Explain why the same sessionKey stays ordered while different sessions can run concurrently (see Queue + lanes).
3. Explain the three safety layers (sandbox / policy / approvals) and what each layer owns (see Sandbox vs tool policy vs elevated).
4. Trace a run from runId to tools and final output using logs (see Logging).