登录
Back to AI tools
Selection guide6 reviewed entries

Agent Frameworks

Compare agent frameworks by planning, tool use, state, workflow orchestration, and human handoff.

For developers and product teams extending a model from one answer into a multi-step task.

How to choose

  • The boundary between deterministic workflows and autonomous planning
  • State, memory, and multi-agent orchestration
  • Permissions, budgets, and stop conditions
  • Debugging, evaluation, and production deployment

DECISION PATH

Start with one real task, inspect the official details, then compare only the candidates that pass your access and quality checks.

Browse all tools

CURATED RESOURCES

Recommended resources

Each entry links to its detail and official source
L

LangChain

构建 LLM 应用、RAG 和 Agent 的开源框架。

GlobalFreeAgent+1

Updated 2026-08-28

Visit official site
L

LangGraph

LangChain 团队的图结构 Agent 编排框架，适合生产。

GlobalFreeAgent+1

Updated 2026-08-28

Visit official site
CrewAI

CrewAI

多 Agent 协作框架，按角色分工完成复杂任务。

OverseasFreemiumMulti-agent+1

Updated 2026-08-28

Visit official site
Dify

Dify

开源 LLM 应用开发平台，可视化编排 Agent 与 RAG。

GlobalFreemiumLLMOps+1

Updated 2026-08-28

Visit official site
Coze

Coze

字节跳动的零代码 Agent 平台，一键发布到多渠道。

DomesticFreemiumNo-code+2

Updated 2026-08-28

Visit official site
mcp-agent

mcp-agent

Build effective agents using Model Context Protocol and simp

GlobalFreeOpen source+1

Updated 2026-09-01

Visit official site

EDITORIAL

Read before choosing

All articles
RAG 与 Agent

Agent 不是自动驾驶：工具调用的边界设计

分析 Agent 的规划、调用、观察循环，以及为什么必须设置权限和预算。

RAG 与 Agent

工作流优先：什么时候不该使用 Agent

比较固定流程、工作流和 Agent，帮助团队避免为了智能而增加不可控复杂度。

AI 编程

AI 应用的评测集怎么建立

从真实任务采样，构建可持续迭代的离线评测集。

行业实践

AI 项目上线前的 20 项检查

一份面向小团队的发布清单，覆盖质量、安全、成本和可回滚性。

Related topics

All topics

Agent

Agent frameworks, tool calling, and workflows.

Workflow Automation

Connect AI to business processes and automate orchestration.

RAG

Retrieval-augmented generation, enterprise knowledge bases, and document Q&A.