引用Google Gemini的释义：
Agentic AI frameworks are software toolkits that provide structure, components, and protocols for building autonomous AI agents that can reason, plan, and act to achieve goals, streamlining development with features like memory, tool integration (APIs), and multi-agent orchestration, with popular examples including AutoGen, CrewAI, LangGraph, and Semantic Kernel, enabling complex automation beyond simple prompts.
Core Components & Capabilities
- Planning & Reasoning: Breaking down tasks, selecting tools, tracking progress.
- Autonomy: Operating independently, adjusting strategies.
- Memory: Retaining context and past actions for better reasoning.
- Tool Integration: Accessing external APIs, databases, and systems.
- Multi-Agent Orchestration: Coordinating multiple agents for complex workflows.
- Human-AI Interaction: Interfaces for feedback and oversight.
这里其实应该先看一下GenAI和AgenticAI的区别
GenAI 解决"生成什么"，Agentic AI 解决"为了目标要做什么、下一步做什么"。
👉 GenAI 是"会写、会说、会生成"的大脑
👉 Agentic AI 是"会思考下一步并动手"的数字员工
GenAI 擅长
-
文本生成（文案、代码、方案）
-
总结、改写、翻译
-
单轮或多轮问答
-
规则内推理（RAG + Prompt）
从本质上看："输入 → 输出"
Agentic AI 擅长
-
拆解目标
-
制定计划
-
调用工具 / 系统
-
多步执行
-
根据结果动态调整策略
从本质上看是："目标 → 行动闭环"
二者在关键技术上的差异有
Agentic AI = GenAI + 规划 + 工具 + 状态 + 决策