← Back to Tutorials

7. Practice & Resources

Quick Reference Guide

TaskOpenAI SDKCrewAILangGraphAutoGen
Create agentAgent(name, instructions)Agent(role, goal)add_node(name, fn)AssistantAgent(name)
Add tool@function_tool@toolInclude in stateregister_function
Multi-agenthandoff()Crew(agents)Multiple nodesGroupChat
MemoryCustom implBuilt-inState + checkpointConversation history
Structured outputoutput_schemaPydanticTypedDict stateCustom models

Framework Selection Guide

Use CaseBest Framework
Simple assistant with toolsOpenAI Agents SDK
Role-based team collaborationCrewAI
Complex stateful workflowsLangGraph
Distributed multi-agent systemsAutoGen
Custom tool ecosystemsMCP + any framework

Useful Resources

6-Week Study Plan

WeekTopicActivities
1Foundation WeekSetup, patterns, multi-LLM, frameworks overview
2OpenAI Agents SDKSDK fundamentals, sales agents, guardrails, deployment
3CrewAIMulti-agent crews, financial research, custom tools
4LangGraphState graphs, chatbots, browser automation, checkpointing
5AutoGenGroup chat, multimodal, distributed runtimes
6MCP + CapstoneMCP servers, trading floor capstone, framework selection

Certificate Path

✏️ Final Capstone: Build an end-to-end multi-agent trading system using at least two different frameworks. Include: (1) a market data retrieval agent, (2) a news sentiment analysis agent, (3) a portfolio management agent, (4) MCP servers for data access, (5) a Gradio UI, and (6) deployment to HuggingFace Spaces.