CrewAI

Status

📋 Todo — send sources when ready.

Learning roadmap

• Mental model — Agents, Tasks, Crew, Process
• Defining an agent (role, goal, backstory, tools, LLM)
• Tasks — description, expected output, agent assignment
• Sequential vs hierarchical process
• Tools — built-in and custom
• Memory (short-term, long-term, entity)
• Delegation between agents
• Output parsing & structured outputs
• CrewAI Flows (newer, more deterministic alternative)
• When to choose CrewAI vs LangGraph vs raw multi-agent code

What I want to be able to do after this topic

• Spin up a multi-agent crew for a concrete task (research → writing → editing pipeline, etc.).
• Equip agents with custom tools.
• Choose between sequential and hierarchical processes.
• Compare CrewAI vs LangGraph honestly when picking a framework.