共 16 条相关资源。
OpenAI 官方轻量 Agent 框架：以 Agent、Handoff（任务移交）、Guardrail（安全护栏）三个原语构建多 Agent 系统，代码量少、心智负担低，是 Swarm 的生产级继任者。OpenAI's lightweight agent framework built on three primitives — agents, handoffs and guardrails — the production-ready successor to Swarm.
Pydantic 团队出品的 Agent 框架：用类型注解约束模型输出，把「结构化输出 + 依赖注入 + 可测试」做到极致，工程化程度高，适合严谨的后端项目。Agent framework from the Pydantic team: typed, validated outputs with dependency injection and testability — built for serious backend engineering.
最流行的 LLM 应用开发框架，提供链式调用、Agent 编排、工具调用与 RAG 全套能力，Python 与 JS 双语言生态，适合快速搭建各类 AI Agent 应用。The most popular framework for building LLM applications: chains, agent orchestration, tool calling and RAG out of the box, in both Python and JavaScript.
基于图（状态机）的 Agent 编排框架，擅长多 Agent 协作、循环与人工介入流程，是 LangChain 官方推荐的复杂 Agent 生产方案。Graph-based agent orchestration from the LangChain team. Ideal for multi-agent collaboration, loops and human-in-the-loop workflows in production.
Google 开源的 Agent Development Kit：原生对接 Gemini 与 Vertex AI，内置多 Agent 层级编排、工具生态与评测能力，适合构建企业级 Agent 服务。Google's open-source Agent Development Kit with native Gemini/Vertex AI integration, hierarchical multi-agent orchestration and built-in evaluation.
HuggingFace 出品的极简 Agent 库：主打 CodeAgent——让模型直接写 Python 代码作为行动（比 JSON 工具调用更灵活），几行代码即可搭建带工具的 Agent。HuggingFace's minimalist agent library featuring CodeAgents that write Python code as actions, more flexible than JSON tool calls.
微软的企业级 LLM 编排 SDK：C#/.NET 与 Python 双支持，插件化技能、Planner 规划与 Azure 深度集成，是 .NET 技术栈做 AI 应用的标准答案。Microsoft's enterprise SDK for LLM orchestration in C#/.NET and Python, with plugins, planners and deep Azure integration.
高性能 Agent 框架（原 phidata）：支持多模态、记忆、知识库与多 Agent 团队编排，强调低开销与生产可用，内置 40+ 预置 Agent 模板。A high-performance agent framework (formerly phidata) supporting multimodal I/O, memory, knowledge and multi-agent teams with low overhead.
阿里巴巴开源的多 Agent 框架：强调消息传递式协作与容错，内置可视化调试（Studio）与分布式部署能力，中文文档完善。Alibaba's multi-agent framework with message-passing collaboration, fault tolerance, a visual Studio and solid Chinese docs.
TypeScript 全栈 Agent 框架：Agent、工作流、工具、记忆、RAG、评估一体化，与 Next.js 生态深度集成，前端团队做 AI 应用的顺手选择。A TypeScript-first agent framework with agents, workflows, tools, memory, RAG and evals integrated — a natural fit for Next.js teams.
MemGPT 团队的进化版 Agent 框架：核心是分层记忆管理（核心记忆/归档记忆自动分页），让 Agent 拥有跨会话长期记忆，支持状态持久化与多 Agent。The evolution of MemGPT: tiered memory management that gives agents long-term memory across sessions, with state persistence and multi-agent support.
引爆 2023 年 Agent 热潮的开山项目，如今已重构为可视化 Agent 构建平台（低代码积木 + 市场），适合搭建自动化任务流程。The project that started the 2023 agent wave, now rebuilt as a visual low-code agent builder with a block marketplace.
把软件公司「角色分工」搬进多 Agent 系统：产品经理、架构师、工程师协作，输入一句话需求输出代码与文档。A multi-agent system simulating a software company: PM, architect and engineer roles turn one-line requirements into code and docs.
微软开源的多 Agent 对话框架，让多个 Agent 通过对话协作完成复杂任务，支持人机混合参与，学术与工业界应用广泛。Microsoft's open-source framework where multiple agents converse to solve complex tasks, with human-in-the-loop support.
原 OpenDevin，开源的 AI 软件工程师 Agent：能改代码、跑命令、浏览网页、提交 PR，在 SWE-bench 上表现领先。Formerly OpenDevin: an open-source AI software engineer that edits code, runs commands, browses and opens PRs.
以「角色扮演」为核心的多 Agent 框架：为每个 Agent 设定角色、目标与工具，像组建团队一样编排 AI 协作，上手简单、社区活跃。Role-playing multi-agent framework: assign roles, goals and tools to each agent and orchestrate them like a team.