← AI Agents & RAG learning pathAI Agents & RAG · Module 7 of 10
AI agents
Build agents that plan, use tools and work in loops.
📖 Key ideas
- An agent is an LLM in a loop: think → act (tool) → observe → repeat.
- Agents handle multi-step tasks like research or coding.
- Memory stores conversation and task state.
- Frameworks: LangGraph, CrewAI, LlamaIndex, OpenAI/Anthropic agent SDKs.
- Keep agents simple; add autonomy only when needed.
💡 Real-life example: An agent is a trainee who plans steps, uses office tools and checks results until the job is done.
AgentLLM that loops with tools toward a goal
ReActReason + Act prompting pattern
MemoryStored context across steps
Stopping conditionWhen the agent ends its loop