AI Agent Frameworks: CrewAI, OpenAI Agents SDK, Claude Agent SDK, ADK, PydanticAI
Coming soon BeginnerIntermediateAdvanced
Get notified on YouTube Udemy certificate: coming soon
- Udemy
- Coming soon
- Lessons
- 30
- Planned length
- About 9 h 4 min
- Projects
- 2
- Practicals
- Planned: a hands-on practical and a Udemy assignment in every section, plus a capstone
What you will be able to do
- Write tool names, descriptions and schemas that work
- Implement ReAct and read its trace
- Separate planning from doing and keep long tasks on track
- Add a critic step that improves answers
- Pick the memory type for a use case
- Recall user facts and relationships across sessions
Syllabus
AI Agents: Tools, Reasoning and Planning
- Design Tools Agents Use Correctly Write tool names, descriptions and schemas that work
- ReAct: Reason + Act Implement ReAct and read its trace
- Plan-and-Execute and Long-Running Agents Separate planning from doing and keep long tasks on track
- Reflection and Self-Critique Add a critic step that improves answers
AI Agents: Agent Memory
- Agent Memory Types: Short, Long, Episodic, Semantic, Procedural Pick the memory type for a use case
- Long-Term and Graph Memory: Vector Stores, Mem0, Letta, Zep Recall user facts and relationships across sessions
- Conversation Memory, Compaction and Forgetting Keep long chats in budget and decide what to forget
AI Agents: Agent Frameworks (One Lesson Each)
- OpenAI Agents SDK Build an agent with handoffs and guardrails
- CrewAI Build a role-based crew of agents
- AutoGen / AG2 (and Microsoft's Agent Stack) Build a conversational multi-agent system
- smolagents Build a code-acting agent in a few lines
- PydanticAI Build a type-safe agent with validated outputs
- Google Agent Development Kit (ADK) Build and run an agent with ADK
- Same Agent, Four Frameworks (+ TypeScript: Mastra, Vercel AI SDK) Compare code size, control, cost and traces
AI Agents: Multi-Agent Patterns
- Why Multi-Agent, and When Not To Know the cost and benefit of more agents
- Supervisor and Orchestrator-Workers Patterns Route and split work across specialist agents
- Swarm and Handoffs Let agents hand off to each other
- Hierarchical Teams, and Why Multi-Agent Systems Fail Build teams of teams and diagnose loops and drift
- Debate and Critique Agents Improve answers with adversarial agents
- Project: Deep Research Agent Team Ship a team that searches, drafts, cites and fact-checks
Frontier and Advanced: Agent Memory Systems
- Why Agents Forget: Memory as a System, Not a Prompt Explain write, store, retrieve and forget for agent memory
- Build Episodic and Semantic Memory From Scratch Store events and facts, retrieve them by relevance and recency
- Mem0 Hands-On: Extract, Update and Scope Memories per User Add per-user long-term memory to an agent
- Letta (MemGPT-Style) Agents: Self-Editing Memory Build an agent that manages its own memory
- Test and Govern Agent Memory: Recall, Staleness, Deletion Prove memory helps, and delete it on request
Frontier and Advanced: MCP and A2A, Advanced
- A2A Advanced: Agent Cards, Tasks, Streaming, Auth Expose an agent over A2A with auth and streaming
- Project: Two Agents, Two Frameworks, One A2A Link Connect agents built in two frameworks through A2A
Frontier and Advanced: Frontier Agent Engineering
- Why Agents Fail in Production, and Service Levels List the failure modes and set targets for an agent
- Agent Reliability: Retries, Idempotency, Checkpoints Make tool calls safe to retry and agent runs resumable
- Durable Long-Running Agents: Resume After a Crash Run an agent for hours and survive restarts
Take before
Lesson list from the ADITEJUS curriculum. Planned lengths are estimates per lesson; Udemy shows the final length.