Agent Frameworks Overview
Frameworks such as LangGraph, CrewAI, and Semantic Kernel provide the plumbing for agent state, tool calls, and control flow.
Prerequisites
Overview
An agent loop can be hand-rolled, but most teams reach for a framework to handle the repetitive plumbing — tracking state across steps, routing between nodes, and retrying failed tool calls — so application code can focus on the actual task logic.
Where It Fits
Your Task Logic
Agent Framework
State, routing, retriesModel
Tools
Key Points
- Graph-based control flow
- LangGraph models an agent as an explicit graph of nodes and edges, making branching and loops visible rather than buried in code.
- Role-based orchestration
- CrewAI organizes multiple agents by role and goal, suited to tasks that naturally split across specialists.
- Enterprise integration
- Semantic Kernel targets .NET/enterprise stacks with first-class plugin and planning support.
Interview Question
What does an agent framework actually give you that hand-rolling an agent loop doesn’t?
Mainly the repetitive plumbing — persisting state across steps, retrying failed tool calls, and making control flow (branches, loops) explicit rather than buried in conditional code. It doesn’t replace the actual task logic or the decision of what tools an agent should have; it just removes the boilerplate around running the loop reliably.
Explain It in 30 Seconds
Agent frameworks like LangGraph, CrewAI, and Semantic Kernel handle the repetitive plumbing of an agent loop — state, routing, retries — so application code can focus on task-specific logic rather than reimplementing that machinery each time.
Real-World Stack
Technologies commonly used to implement this in production.