Agentic Architecture & Orchestration

Design agentic systems with the Claude API and Agent SDK: loops, orchestration, guardrails and state.

1. 1.1Agentic Loops and Stop ReasonsHow a Claude agent actually runs, request by request, and why the stop_reason field decides what your code does next.9 min
2. 1.2Orchestration Patterns: Chains, Routers and Orchestrator-WorkersThe standard workflow and agent patterns for composing Claude calls, and how to pick the simplest one that fits.9 min
3. 1.3Subagent Invocation and Context PassingWhat a subagent is, what it does and does not inherit from its parent, and how to pass context across the boundary deliberately.9 min
4. 1.4Workflow Enforcement, Guardrails and HandoffWhy prompts are suggestions and hooks are rules, and how to design guardrails and clean hand-offs in agentic systems.9 min
5. 1.5Agent SDK Hooks and Lifecycle EventsHow the Claude Agent SDK exposes the Claude Code harness as a library, and how hooks let your code observe and steer every stage of a run.9 min
6. 1.6Task Decomposition StrategiesHow to split a large task into pieces an agent can complete reliably, and how to decide the granularity, ordering and verification of each piece.8 min
7. 1.7Session State, Persistence and ResumptionWhere state lives in a Claude system, from the stateless API to Claude Code sessions, the memory tool and persistent project memory, and how to resume safely.9 min