Architectures and Design Principles for Agentic AI Systems
Agentic AI arrives with a wall of words.
Agents, harnesses, orchestrators, MCP, A2A, ReAct, supervisors, frameworks… The terminology grows faster than the ideas underneath it. This lesson takes a different route through it.
AgentAgentic AIWorkflowOrchestratorHarnessToolFunction CallingMCPA2AMemoryStateContextRAGPlannerRouterSupervisorReflectionReActHandoffMulti-AgentGuardrailsFrameworkLangGraphStrands AgentsAgents SDKObservabilityEvaluation
Do we need to memorise all of these before we can design Agentic AI systems?