Learn to design and build AI agents that plan, reason, and take actions — from single-task bots to multi-agent systems orchestrating complex workflows. This course covers agent architectures like ReAct and Plan-and-Execute, tool use and function calling, frameworks such as LangGraph and AutoGen, memory systems, and multi-agent coordination. Throughout, it treats reliability honestly: you will learn guardrails, permissions, observability, and human-in-the-loop design, because autonomy without oversight is how agent projects fail.

Modules

Agent Architectures and Reasoning Patterns

Lessons

• What Turns a Language Model into an Agent Preview
• The ReAct Pattern of Reasoning and Acting
• Plan-and-Execute and Choosing Your Architecture

Tool Use and Function Calling

Lessons

• How Function Calling Actually Works
• Designing Tool Schemas Agents Can Use Reliably
• Handling Tool Errors and Unreliable Calls

Building Agents with LangGraph and AutoGen

Lessons

• Modeling Agent Workflows as Graphs in LangGraph
• Conversation-Driven Agents with AutoGen
• Testing and Debugging Agent Systems

Memory Systems for Agents

Lessons

• Short-Term Memory and the Context Window Budget
• Long-Term Memory with Retrieval and Stores
• Episodic Memory and Learning from Past Runs

Multi-Agent Orchestration

Lessons

• When Multiple Agents Beat One
• Orchestration Patterns from Supervisors to Swarms
• Agent Communication and Shared State

Guardrails Oversight and Production Deployment

Lessons

• Why Agents Fail and What Reliability Really Means
• Permissions Guardrails and Least-Privilege Actions
• Human-in-the-Loop Design and Production Observability