## TL;DR
AI agent frameworks provide the orchestration layer for building autonomous AI applications. LangChain pioneered composable LLM pipelines; AutoGen introduced multi-agent conversations; CrewAI enables role-based agent teams.

## Core Explanation
Framework comparison: LangChain — most mature, broadest integration ecosystem (500+ tools/APIs). LangGraph — stateful, graph-based agent control flow. AutoGen — multi-agent with structured message passing. CrewAI — role-based delegation (manager, researcher, writer). Semantic Kernel (Microsoft) — enterprise-grade, native .NET/Python integration.

## Detailed Analysis
The ReAct pattern (Reasoning + Acting) is the dominant agent loop: think → act → observe → think. Tool calling (function calling) enables agents to invoke APIs, databases, and code. Memory systems (short-term via conversation history, long-term via vector stores) persist context across interactions.

## Further Reading
- LangChain Documentation
- Microsoft AutoGen GitHub
- Anthropic: Tool Use Guide