Chapter Studio
AI Agents & Tool Calling
Follow an agent as it looks up orders, reads policies, and prepares a return draft. Then connect tools and run your own configuration with Luma Shop sample data.
What is an agent?
Compare a chatbot, a fixed workflow, and an agent without hiding where the decisions come from.
1 focused lesson02Tool calling
Follow a structured request through argument validation, application execution, and a matched result.
1 focused lesson03The agent loop
Trace a complete return task and see when the next step is another call, a question, or an answer.
1 focused lesson04State & memory
Separate current context, conversation state, and information explicitly saved for later retrieval.
2 short lessons05Reliability
Validate tool requests, stop unproductive loops, and prevent a retry from creating duplicate drafts.
3 short lessons06Agentic RAG
Make retrieval a tool choice, inspect actual returned passages, and keep unsupported claims out of the answer.
1 focused lesson07Agent Builder
Configure one simulated agent with built-in tools, guided tasks, context, and explicit execution limits.
1 focused lessonRuns offline. Agent decisions are authored simulations. Tool calls execute against sample data.