AI Agent 开源项目大全：2026 年最佳智能体工具推荐

Superpowers

Agentic 技能框架与软件开发方法论，通过子代理驱动开发模式，提供可复用的 AI 技能组合。

affaan-m/ECC

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond

Hermes Agent

NousResearch 开源的长期运行 AI agent，可执行本地工具、定时任务和持续自动化维护流程。

The agent that grows with you

deepseek-ai/deepseek-harness

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：DeepSeek Harness: Everything is a Plugin.

DeepSeek Harness: Everything is a Plugin.

OpenCode

开源编码智能体，支持自然语言驱动的代码编写、调试与项目管理，可在终端中直接与 AI 协作完成开发任务。

Significant-Gravitas/AutoGPT

经典自主 AI Agent 平台，支持目标驱动任务拆解、工具调用与可扩展 Agent 工作流，适合研究自主智能体产品形态与多步骤任务自动化。

AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.

microsoft/markitdown

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Python tool for converting files and office documents to Markdown

Python tool for converting files and office documents to Markdown

Ollama

本地大语言模型一键部署框架，支持 Kimi-K2.6、GLM-5.1、Qwen、DeepSeek 等主流开源模型的本地运行与服务化。

firecrawl/firecrawl

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：The API to search, scrape, and interact with the web at scale. 🔥

The API to search, scrape, and interact with the web at scale. 🔥

langgenius/dify

生产级 Agentic Workflow 开发平台，面向 RAG、工具调用、编排和低代码 AI 应用搭建。

Production-ready platform for agentic workflow development.

Langflow

可视化 AI 智能体与工作流构建平台，支持拖拽式编排多 Agent 协作、RAG 管道和 LLM 应用部署。

Agency Agents

完整 AI 智能体代理集合，涵盖前端开发、社区运营、创意生成等专业角色，每个智能体具备独立人格、流程和可交付成果。

LangChain

面向 LLM 应用与智能体工程的核心框架生态，适合编排工具调用、RAG 和复杂 AI 工作流。

Agent engineering framework and ecosystem for building LLM applications.

Claude Code

Anthropic 官方终端 AI 编程智能体，直接在命令行中理解代码库、执行开发任务、处理 git 工作流，通过自然语言指令加速开发。

farion1231/cc-switch

这是一个 AI 智能体 方向的 Rust 开源项目，主要用于：A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Gemini CLI & Hermes Agent. Only official website: ccswitch.io

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Gemini CLI & Hermes Agent. Only official website: ccswitch.io

OpenAI Codex CLI

OpenAI 轻量级终端编码智能体，在命令行中运行，支持自然语言驱动的代码编写与项目管理。

Browser Use

让 AI agents 能操作网页，适合浏览器自动化和在线任务执行。

Make websites accessible to AI agents for browser automation and web task execution.

Gemini CLI

Google 开源终端 AI 智能体，集成 Gemini 模型能力，在命令行中提供代码生成、理解和自动化能力。

TradingAgents

多 Agent LLM 金融交易框架，整合多个专业智能体进行市场分析、策略制定与风险管理，适用于量化交易研究。

punkpeye/awesome-mcp-servers

高活跃 MCP Server 资源合集，集中整理可接入 Claude、Cursor、ChatGPT 与各类 Agent 的工具服务器，适合快速寻找数据库、浏览器、文件、API 等外部能力。

A collection of MCP servers.

Claude-Mem

跨 Agent 持久化上下文记忆引擎，自动捕获并压缩智能体会话信息，在后续会话中注入相关上下文，兼容 Claude Code、Codex、Gemini CLI 等主流 Agent。

RAGFlow

基于深度文档理解的开源 RAG 引擎，融合检索增强生成与智能体能力，为 LLM 提供高质量上下文层。

modelcontextprotocol/servers

Model Context Protocol 官方参考 MCP Server 集合，提供文件系统、数据库、浏览器等标准工具服务，适合给 Agent 接官方 MCP 能力。

Model Context Protocol Servers

OpenHands

开源自主 AI 软件开发智能体，可理解代码库、执行开发任务、处理 bug 修复与功能实现，支持多模型切换。

bytedance/deer-flow

字节开源的长程任务 SuperAgent 框架，围绕研究、编码、内容生成、沙箱、记忆和子智能体协作，适合分钟到小时级复杂任务。

An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.

Panniantong/Agent-Reach

为 AI Agent 提供跨 Twitter、Reddit、YouTube、GitHub、B 站、小红书等平台的读取与搜索能力，适合研究和信息收集场景。

Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.

shareAI-lab/learn-claude-code

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1

Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1

ComposioHQ/awesome-claude-skills

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：A curated list of awesome Claude Skills, resources, and tools for customizing Claude AI workflows

A curated list of awesome Claude Skills, resources, and tools for customizing Claude AI workflows

ruvnet/ruflo

多智能体 meta-harness，用于编排自主 swarm、协调并行任务和部署智能体集群，适合探索企业级 multi-agent 编排与协作。

🌊 The leading agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated

headroomlabs-ai/headroom

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 60-95% fewer tokens, same answers. Library, proxy, MCP server

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 60-95% fewer tokens, same answers. Library, proxy, MCP server

santifer/career-ops

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：AI-powered job search system built on Claude Code. 14 skill modes, Go dashboard, PDF generation, batch processing

AI-powered job search system built on Claude Code. 14 skill modes, Go dashboard, PDF generation, batch processing

FoundationAgents/MetaGPT

多智能体软件开发框架，把产品经理、架构师、工程师等角色拆成协作 Agent，适合研究自然语言到软件原型的自动化流程。

🌟 The Multi-Agent Framework: First AI Software Company, Towards Natural Language Programming

code-yeongyu/oh-my-openagent

面向 token 效率的编码智能体 harness，强调工具编排、多代理协作与可扩展技能，适合研究 coding agent 运行时与工作流优化。

omo/lazycodex: The coding agent for tokenmaxxers;the one and only agent harness for complex codebases. For your Codex, for your OpenCode

stablyai/orca

面向并行编码智能体的新一代 IDE，可把多个 coding agent 作为协作队列管理，适合探索多 Agent 开发、任务分派和移动端查看流程。

Orca is the next-gen IDE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop and mobile.

docling-project/docling

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Get your documents ready for gen AI

Get your documents ready for gen AI

Mintplex-Labs/anything-llm

本地优先的 AI 工作区与 Agent 平台，支持私有知识库、工具调用和多模型接入，适合自托管智能助手与企业内部 Agent 场景。

Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience

Mem0

面向 AI agents 的通用记忆层，让智能体具备个性化和状态保持能力。

Universal memory layer for AI agents, enabling personalized and stateful agent experiences.

ZhuLinsen/daily_stock_analysis

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：LLM 驱动的多市场股票智能分析系统：多源行情、实时新闻、决策看板与自动推送，支持零成本定时运行。 LLM-powered multi-market stock analysis system with multi-source market data, real-time news, decision dashboard, automated notif

LLM 驱动的多市场股票智能分析系统：多源行情、实时新闻、决策看板与自动推送，支持零成本定时运行。 LLM-powered multi-market stock analysis system with multi-source market data, real-time news, decision dashboard, automated notif

sansan0/TrendRadar

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载，你的 AI 舆情监控助手与热点筛选工具！聚合多平台热点 + RSS 订阅，支持关键词精准筛选。AI 智能筛选新闻 + AI 翻译 + AI 分析

⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载，你的 AI 舆情监控助手与热点筛选工具！聚合多平台热点 + RSS 订阅，支持关键词精准筛选。AI 智能筛选新闻 + AI 翻译 + AI 分析

AutoGen

微软开源的多智能体应用框架，用于自动化 AI 对话和协作。

Microsoft's framework for multi-agent AI applications and conversations.

MemPalace/mempalace

开源 AI 记忆系统，面向 Agent 的长期记忆、检索和上下文管理，可作为 MCP/RAG 工作流里的记忆层参考。

The best-benchmarked open-source AI memory system. And it's free.

CrewAI

基于角色分工的智能体协作框架，适合自动化团队式任务流程。

Role-based agent framework for automating collaborative AI workflows.

crewAI

角色扮演型多 Agent 编排框架，通过定义 Agent 角色、任务和工具实现自主协作，支持复杂工作流的自动化。

hugohe3/ppt-master

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：AI generates a real, editable PowerPoint from any document — native shapes & animations, speaker notes voiced as audio narration, and the option to follow your own .pptx template,

AI generates a real, editable PowerPoint from any document — native shapes & animations, speaker notes voiced as audio narration, and the option to follow your own .pptx template,

dockur/windows

这是一个 AI 智能体 方向的 Shell 开源项目，主要用于：Windows inside a Docker container

Windows inside a Docker container

LlamaIndex

面向私有数据、知识库和文档智能体的数据框架，适合构建 RAG、检索和企业知识助手。

Data framework for connecting private or domain data to LLM apps and agents.

CherryHQ/cherry-studio

面向多模型对话、智能助手和自主 Agent 的桌面生产力工作台，适合本地管理常用模型、助手和工具流。

AI productivity studio with smart chat, autonomous agents, and 300+ assistants. Unified access to frontier LLMs

Chrome DevTools MCP

Chrome DevTools 官方 MCP 服务，让 AI 编程智能体能够连接浏览器调试、检查页面和分析前端问题。

Official Chrome DevTools MCP server that gives coding agents browser debugging capabilities.

HKUDS/nanobot

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Lightweight, open-source AI agent for your tools, chats, and workflows

Lightweight, open-source AI agent for your tools, chats, and workflows

zhayujie/CowAgent

轻量开源 AI 助手与 Agent Harness，支持任务规划、工具/技能调用、记忆和多渠道接入。

Open-source super AI assistant & Agent Harness. Plans tasks, runs tools and skills, autonomously grows with memory and knowledge. Multi-model, multi-channel. Lightweight, extensible, one-line install (formerly chatgpt-on-wechat).

sickn33/antigravity-awesome-skills

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Installable GitHub library of 1,600+ agentic skills for Claude Code, Cursor, Codex CLI, Gemini CLI, Antigravity, and more. Includes specialized plugins, installer CLI, bundles, wor

Installable GitHub library of 1,600+ agentic skills for Claude Code, Cursor, Codex CLI, Gemini CLI, Antigravity, and more. Includes specialized plugins, installer CLI, bundles, wor

vercel-labs/agent-browser

面向 AI Agent 的浏览器自动化 CLI，可让智能体在网页环境中执行浏览、操作和任务自动化。

Browser automation CLI for AI agents

Agno

智能体平台框架，用于构建、运行和管理生产级 AI agent 系统。

Agent platform framework for building, running, and managing production AI agent systems.

Phidata

Python 原生 AI Agent 开发框架，内置知识库、工具和模型抽象，快速构建生产级 Agent 应用。

LangGraph

基于图结构编排有状态、多智能体工作流，适合复杂 AI 应用。

Build stateful, multi-agent workflows with graph-based orchestration.

AstrBot

集成多个 IM 平台、LLM、插件和 AI 能力的智能体助手开发框架，适合搭建可扩展的聊天机器人与自动化助手。

AI agent assistant and development framework with IM platform integrations, plugins and LLM features.

wshobson/agents

Multi-harness agentic plugin marketplace for Claude Code, Codex CLI, Cursor, OpenCode, GitHub Copilot, and Gemini CLI

UI-TARS Desktop

字节开源的多模态 AI Agent 桌面栈，面向 GUI 操作、Computer Use、浏览器任务和 MCP 集成，适合研究桌面/浏览器代理能力。

Open-source multimodal AI agent stack for computer-use, GUI operation and browser automation.

CopilotKit

面向 React / Angular 的前端智能体与生成式 UI 框架，适合把 AI 助手、可交互 Copilot 和 AG-UI 协议接入产品界面。

Frontend stack for AI agents and generative UI in React and Angular, and the AG-UI protocol.

khoj-ai/khoj

可自托管的 AI 第二大脑，支持从个人文档和网页中检索问答、构建自定义 Agent 与定时自动化，适合知识管理和研究助手。

Your AI second brain. Self-hostable. Get answers from the web or your docs. Build custom agents, schedule automations, do deep research. Turn any online or local LLM into your personal, autonomous AI (gpt, claude, gemini

Playwright MCP

微软 Playwright MCP 服务，把浏览器自动化、测试和页面操作能力开放给 AI 智能体使用。

Microsoft Playwright MCP server for browser automation and testing through AI agents.

OpenBMB/ChatDev

用多个 LLM 角色模拟软件公司协作，从需求、设计到代码生成组织流程，适合对比多 Agent 协同开发模式。

ChatDev 2.0: Dev All through LLM-powered Multi-Agent Collaboration

GitHub MCP Server

GitHub 官方 MCP 服务，让 Claude、Cursor 等智能体安全访问仓库、Issue、PR 和代码工作流。

GitHub official MCP server for repository, issue, pull request and code workflows.

zeroclaw-labs/zeroclaw

轻量级全自主 AI 个人助手基础设施，支持任意操作系统与设备，可自主执行复杂任务链。

Fast, small, and fully autonomous AI personal assistant infrastructure, any OS, any device.

googleworkspace/cli

Google Workspace CLI — one command-line tool for Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, and more. Dynamically built from Google Discovery Service. Includes AI agent skills.

ScrapeGraphAI/Scrapegraph-ai

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Python scraper based on AI

Python scraper based on AI

can1357/oh-my-pi

终端 AI 编码智能体，支持哈希锚定编辑、工具 harness、LSP、Python、浏览器与子代理，适合命令行重度开发与 agentic coding。

⌥ AI Coding agent for the terminal — hash-anchored edits, optimized tool harness, LSP, Python, browser, subagents, and more

supermemoryai/supermemory

这是一个 后端数据库 方向的 TypeScript 开源项目，主要用于：Memory and context engine + app that is extremely fast, scalable, and can be run fully locally. The Memory API for the AI era

Memory and context engine + app that is extremely fast, scalable, and can be run fully locally. The Memory API for the AI era

simstudioai/sim

用于构建、部署和编排 AI Agent 的工作流平台，适合把模型、工具和业务自动化流程连接成可运行的团队级 Agent 系统。

Build, deploy, and orchestrate AI agents. Sim is the central intelligence layer for your AI workforce.

GPT Researcher

自主研究智能体，可围绕主题检索资料、汇总来源并生成结构化深度研究报告。

Autonomous research agent that gathers sources and produces long-form research reports.

OpenAI Agents SDK

OpenAI 官方 Python 智能体 SDK，支持工具调用、任务交接和安全护栏。

Official Python SDK for building agentic apps with tools, handoffs and guardrails.

smolagents

Hugging Face 的轻量智能体库，适合代码执行型 agents 和工具调用流程。

Hugging Face lightweight library for code-thinking agents and tool-using workflows.

Semantic Kernel

微软开源 SDK，用于把 LLM、插件、规划器和 agent 工作流集成到应用。

Microsoft SDK for integrating LLMs, planners, plugins and agentic workflows into applications.

mastra-ai/mastra

面向 TypeScript/Node 生态的 AI 应用与 Agent 框架，提供工作流、工具调用和 MCP 相关能力，适合构建现代 Web 技术栈上的智能体产品。

From the team behind Gatsby, Mastra is a framework for building AI-powered applications and agents with a modern TypeScript stack.

Mastra

来自 Gatsby 团队的 TypeScript AI 应用与 Agent 框架，支持工作流、工具调用、MCP 集成和可观测性，适合构建面向产品的智能体应用。

TypeScript framework for building AI applications and agents with workflows, tools and observability.

FastMCP

Python 风格的 MCP 服务端/客户端开发框架，适合快速把内部工具封装成智能体可调用能力。

Pythonic framework for building MCP servers and clients quickly.

Fosowl/agenticSeek

完全本地化的自主 AI 智能体，无需 API 调用和订阅费用，可在本地运行类似 Manus 的自动化体验。

Fully Local Manus AI. No APIs, No $200 monthly bills. Enjoy an autonomous agent experience for free.

h4ckf0r0day/obscura

面向 AI Agent 的无头浏览器引擎，专为网页自动化与数据采集优化，支持代理级并发操作。

The headless browser for AI agents and web scraping.

Haystack

面向上下文工程、RAG、多模态检索和智能体工作流的 AI 编排框架，适合需要可控管线、路由和记忆的生产级 LLM 应用。

AI orchestration framework for context-engineered LLM apps, RAG pipelines and agent workflows.

zai-org/Open-AutoGLM

开源手机 AI Agent 模型与框架，可自动操控手机完成复杂任务，面向移动端自动化场景。

An Open Phone Agent Model & Framework. Unlocking the AI Phone for Everyone.

enescingoz/awesome-n8n-templates

这是一个 AI 智能体 方向的 开源 开源项目，主要用于：280+ free n8n automation templates — ready-to-use workflows for Gmail, Telegram, Slack, Discord, WhatsApp, Google Drive, Notion, OpenAI, and more. AI agents, RAG chatbots, email

280+ free n8n automation templates — ready-to-use workflows for Gmail, Telegram, Slack, Discord, WhatsApp, Google Drive, Notion, OpenAI, and more. AI agents, RAG chatbots, email

Letta

带长期记忆和自我改进能力的状态型 agent 框架与平台。

Framework and platform for stateful agents with long-term memory and self-improvement loops.

Activepieces

开源 AI 工作流与自动化平台，内置 agents、MCP 集成和大量应用连接器，适合搭建业务自动化。

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

Stagehand

Browserbase 出品的浏览器智能体 SDK，把 Playwright/Puppeteer 自动化与自然语言指令结合，适合网页操作、测试和浏览器 Agent 原型。

The SDK For Browser Agents

floci-io/floci

这是一个 AI 智能体 方向的 Java 开源项目，主要用于：Light, fluffy, and always free - The AWS Local Emulator alternative

Light, fluffy, and always free - The AWS Local Emulator alternative

trycua/cua

面向 Computer-Use Agents 的开源基础设施，提供沙箱、SDK 和评测组件，用于训练与验证能操作桌面的智能体。

Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows).

Tencent/WeKnora

这是一个 AI 智能体 方向的 Go 开源项目，主要用于：Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki

Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki

mksglu/context-mode

面向 AI 编码智能体的上下文窗口优化工具，通过隔离和压缩工具输出减少噪声，适合长任务、MCP 工具链和多平台 coding agent 工作流。

Context window optimization for AI coding agents. Sandboxes tool output, 98% reduction. 15 platforms.

openai/swarm

OpenAI 推出的轻量级多智能体编排教育框架，探索多 Agent 协作的工程化最佳实践。

Educational framework exploring ergonomic, lightweight multi-agent orchestration.

guillaumemeyer/watermarks-remover

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：A privacy-first app that strips AI watermarks from content you own.

A privacy-first app that strips AI watermarks from content you own.

coze-dev/coze-studio

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：An AI agent development platform with all-in-one visual tools, simplifying agent creation, debugging, and deployment like never before. Coze your way to AI Agent creation

An AI agent development platform with all-in-one visual tools, simplifying agent creation, debugging, and deployment like never before. Coze your way to AI Agent creation

Google ADK

Google Agent Development Kit，代码优先的智能体开发、评估和部署工具包。

Google Agent Development Kit: code-first toolkit for building, evaluating and deploying AI agents.

jnMetaCode/agency-agents-zh

这是一个 AI 智能体 方向的 Shell 开源项目，主要用于：🎭 266 个即插即用的 AI 专家角色 — 支持 Hermes Agent/Claude Code/Cursor/Copilot 等 18 种工具，覆盖工程/设计/营销/金融等 20 个部门。含 50 个中国市场原创智能体（小红书/抖音/微信/飞书/钉钉等）。搭配编排器 agency-orchestrator，一句话即可让多位专家按 DAG 自动协作。

🎭 266 个即插即用的 AI 专家角色 — 支持 Hermes Agent/Claude Code/Cursor/Copilot 等 18 种工具，覆盖工程/设计/营销/金融等 20 个部门。含 50 个中国市场原创智能体（小红书/抖音/微信/飞书/钉钉等）。搭配编排器 agency-orchestrator，一句话即可让多位专家按 DAG 自动协作。

Alibaba-NLP/DeepResearch

通义团队开源的 Deep Research Agent，面向长链路资料检索、分析和报告生成，适合关注研究型智能体与中文生态的开发者。

Tongyi Deep Research, the Leading Open-source Deep Research Agent

pydantic/pydantic-ai

Pydantic 团队推出的 AI Agent 框架，用类型校验和结构化输出组织工具调用、模型交互与生产级智能体应用开发。

AI Agent Framework, the Pydantic way

Pydantic AI

Pydantic 团队推出的 Python 智能体框架，强调类型安全、工具调用和模型无关设计。

Python agent framework from the Pydantic team with typed dependencies, tools and model-agnostic design.

liyupi/ai-guide

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：程序员鱼皮的 AI 资源大全 + Vibe Coding 零基础教程，分享 OpenClaw 保姆级教程、大模型玩法（DeepSeek / GPT / Gemini / Claude / GLM）、最新 AI 资讯、Prompt 提示词大全、AI 知识百科（Agent Skills / RAG / MCP / A2A）、AI 编程教程（Harness Eng

程序员鱼皮的 AI 资源大全 + Vibe Coding 零基础教程，分享 OpenClaw 保姆级教程、大模型玩法（DeepSeek / GPT / Gemini / Claude / GLM）、最新 AI 资讯、Prompt 提示词大全、AI 知识百科（Agent Skills / RAG / MCP / A2A）、AI 编程教程（Harness Eng

dzhng/deep-research

开源深度研究助手，通过迭代搜索、阅读和整理信息完成主题研究，适合学习 research agent 的规划、信息收集和报告生成流程。

An AI-powered research assistant that performs iterative, deep research on any topic by combining search engines, web scraping, and large language models. The goal of this repo is to provide the simplest implementation of a deep research agent - e.g. an agent that can refine its research direction overtime and deep dive into a topic.

datawhalechina/easy-vibe

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：💻 vibe coding 2026 | Your First Modern Coding course beginners to master step by step

💻 vibe coding 2026 | Your First Modern Coding course beginners to master step by step

Eliza

开源 agentic operating system，用于构建自治聊天机器人、社交/社区代理和可扩展的多平台智能体运行时。

Open-source agentic operating system for building autonomous chatbots and multi-platform agents.

teng-lin/notebooklm-py

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Unofficial Python API and agentic skill for Google NotebookLM. Full programmatic access to NotebookLM's features—including capabilities the web UI doesn't expose—via Python, CLI, a

Unofficial Python API and agentic skill for Google NotebookLM. Full programmatic access to NotebookLM's features—including capabilities the web UI doesn't expose—via Python, CLI, a

agent0ai/agent-zero

可自托管的 AI Agent 框架，强调长期运行、工具调用、任务执行和可扩展的自动化工作流。

Agent Zero AI framework

C4illin/ConvertX

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：💾 Self-hosted online file converter. Supports 1000+ formats ⚙️

💾 Self-hosted online file converter. Supports 1000+ formats ⚙️

iii-hq/iii

这是一个 AI 智能体 方向的 Rust 开源项目，主要用于：Effortlessly compose, extend, and observe every service in real-time for the first time ever

Effortlessly compose, extend, and observe every service in real-time for the first time ever

arc53/DocsGPT

私有 AI 知识平台，支持 Agent Builder、深度研究、文档分析、多模型接入和企业搜索场景。

Private AI platform for agents, assistants and enterprise search.

AsyncFuncAI/deepwiki-open

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Open Source DeepWiki: AI-Powered Wiki Generator for GitHub/Gitlab/Bitbucket Repositories. Join the discord: https://discord.gg/gMwThUMeme

Open Source DeepWiki: AI-Powered Wiki Generator for GitHub/Gitlab/Bitbucket Repositories. Join the discord: https://discord.gg/gMwThUMeme

CAMEL

面向多智能体沟通、仿真和规模化研究的开源 agent 框架。

Multi-agent framework focused on communicative agents, simulations and agent scaling research.

EvoLinkAI/awesome-gpt-image-2-API-and-Prompts

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：GPT-Image-2 API and Prompts

GPT-Image-2 API and Prompts

MCP Toolbox for Databases

Google 开源的数据库 MCP Server 工具箱，连接 MySQL、Redis、BigQuery、Elasticsearch 等数据源，让 AI Agent 能以 MCP 方式安全访问数据库工具。

Open-source MCP server toolbox that lets agents work with databases through a controlled tool layer.

googleapis/mcp-toolbox

MCP Toolbox for Databases is an open source MCP server for databases.

browser-use/web-ui

browser-use 的可视化 Web UI，让用户在浏览器中运行和调试网页自动化 AI Agent。

🖥️ Run AI Agent in your browser.

GLips/Figma-Context-MCP

为 Cursor 等 AI 编程智能体提供 Figma 布局上下文的 MCP 服务，适合设计稿理解、前端还原和设计到代码流程。

MCP server to provide Figma layout information to AI coding agents like Cursor

n8n-io/self-hosted-ai-starter-kit

这是一个 AI 智能体 方向的 开源 开源项目，主要用于：The Self-hosted AI Starter Kit is an open-source template that quickly sets up a local AI environment. Curated by n8n, it provides essential tools for creating secure, self-hosted

The Self-hosted AI Starter Kit is an open-source template that quickly sets up a local AI environment. Curated by n8n, it provides essential tools for creating secure, self-hosted

OpenSandbox

阿里巴巴开源的安全、快速、可扩展 sandbox runtime，用于给 AI agents 提供隔离执行环境。

Secure, Fast, and Extensible Sandbox runtime for AI agents.

opensandbox-group/OpenSandbox

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Secure, Fast, and Extensible Sandbox runtime for AI agents

Secure, Fast, and Extensible Sandbox runtime for AI agents

yusufkaraaslan/Skill_Seekers

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Convert documentation websites, GitHub repositories, and PDFs into Claude AI skills with automatic conflict detection

Convert documentation websites, GitHub repositories, and PDFs into Claude AI skills with automatic conflict detection

microsoft/RD-Agent

微软开源的研发自动化智能体，聚焦数据驱动的 AI 研发流程，可辅助数据、模型和实验迭代等高价值研发环节。

Research and development (R&D) is crucial for the enhancement of industrial productivity, especially in the AI era, where the core aspects of R&D are mainly focused on data and models. We are committed to automating these high-value generic R&D processes through R&D-Agent, which lets AI drive data-driven AI. 🔗https://aka.ms/RD-Agent-Tech-Report

livekit/agents

LiveKit 的实时语音 AI Agent 框架，适合构建语音助手、实时通话智能体、视频/音频交互和多模态客服场景。

A framework for building realtime voice AI agents.

nanobrowser/nanobrowser

开源 Chrome 浏览器扩展，内置多 Agent 工作流引擎，可在浏览器中自动完成网页操作与数据处理。

Open-Source Chrome extension for AI-powered web automation. Run multi-agent workflows in your browser.

e2b-dev/E2B

为企业级 AI 智能体提供开源安全运行环境和真实工具沙箱，适合需要执行代码、文件操作、外部工具调用的 Agent 应用。

Open-source, secure environment with real-world tools for enterprise-grade agents.

browseros-ai/BrowserOS

开源 Agentic Browser，面向网页浏览、任务执行和浏览器内智能体体验，适合关注 Browser Agent 与个人自动化的开发者。

🌐 The open-source Agentic browser; alternative to ChatGPT Atlas, Perplexity Comet, Dia.

microsoft/agent-framework

微软开源的 AI Agent 构建与编排框架，支持 Python 与 .NET，用于开发、评估和部署多智能体工作流，适合企业级 Agent 系统选型。

A framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Python and .NET.

Microsoft Agent Framework

微软开源的 Python/.NET AI Agent 框架，用于构建、编排和部署单智能体、多智能体工作流，适合企业级 Agent 应用和官方生态选型。

A framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Python and .NET.

Open Deep Research

开源深度研究 agent 示例，围绕搜索、综合分析和报告生成构建。

Open-source deep research agent example built around search, synthesis and report generation.

hangwin/mcp-chrome

基于 Chrome 扩展的 MCP Server，把浏览器标签、页面内容和操作能力暴露给支持 MCP 的 AI 助手，适合浏览器 Agent 与网页自动化场景。

Chrome MCP Server is a Chrome extension-based Model Context Protocol (MCP) server that exposes your Chrome browser functionality to AI assistants like Claude, enabling complex browser automation, content analysis, and se

simular-ai/Agent-S

像人类一样操作电脑的开源 Agent 框架，覆盖 GUI 理解与桌面任务自动化，适合 computer-use、桌面代理与人机协作研究。

Agent S: an open agentic framework that uses computers like a human

SmolAgents

Hugging Face 推出的轻量级 AI Agent 框架，支持多模型、多工具的智能体编排。

Freebuff

一个通过 npm 分发的免费/广告补贴型命令行 AI 编程助手，关联 Codebuff 项目，适合代码阅读、修改和项目辅助开发。

A free/ad-supported command-line AI coding agent distributed via npm, related to Codebuff and positioned for code editing and project assistance.

The-Pocket/PocketFlow

极简 LLM/Agent 工作流框架，用少量代码表达节点、流程和智能体协作，适合想快速理解 Agent 框架核心结构或搭建轻量原型的开发者。

Pocket Flow: 100-line LLM framework. Let Agents build Agents!

EKKOLearnAI/hermes-web-ui

Hermes Agent 的 Web 控制台，聚合多平台 AI 对话、会话管理、定时任务和使用分析，适合需要可视化运营 Agent 的团队参考。

Web dashboard for Hermes Agent — multi-platform AI chat, session management, scheduled jobs, usage analytics

modelcontextprotocol/inspector

Model Context Protocol 官方生态的 MCP Server 可视化测试工具，用于检查工具、资源、提示词和协议响应，适合开发者调试自己的 MCP 服务。

Visual testing tool for MCP servers

alibaba/spring-ai-alibaba

这是一个 AI 智能体 方向的 Java 开源项目，主要用于：Agentic AI Framework for Java Developers

Agentic AI Framework for Java Developers

mcp-use

面向 ChatGPT、Claude 和 AI Agent 的全栈 MCP 框架，可用于开发 MCP 应用、MCP Server 与智能体工具接入流程。

Full-stack MCP framework for developing MCP Apps and MCP Servers for AI agents.

VoltAgent

开源 TypeScript AI agent 工程平台，覆盖智能体编排、工具调用、可观测性和应用集成。

AI Agent Engineering Platform built on an Open Source TypeScript AI Agent Framework

trycompai/crm

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：Comp AI CRM is an open source, CRM designed for AI agents. Agentic-first CRM.

Comp AI CRM is an open source, CRM designed for AI agents. Agentic-first CRM.

microsoft/magentic-ui

微软实验性智能体界面，可让 Agent 在浏览器与本地文件系统之间执行任务，适合观察人机协作、浏览器自动化和安全边界设计。

An experimental agent that works across the browser and local file system

awslabs/mcp

AWS 官方实验室维护的 MCP Servers 集合，把 AWS 云服务能力通过 MCP 暴露给 AI Agent，适合云资源查询、运维自动化和企业 Agent 工具链集成。

Open source MCP Servers for AWS

wonderwhy-er/DesktopCommanderMCP

为 Claude 等智能体提供终端控制、文件系统搜索和差异编辑能力的桌面 MCP 服务，适合本地自动化与开发辅助。

MCP server for Claude with terminal control, file system search and diff file editing capabilities

MervinPraison/PraisonAI

用于快速构建可研究、规划、编码和执行任务的 AI Agent 工作流，内置记忆、RAG 和多模型支持。

Hire a 24/7 AI Workforce with autonomous agents that research, plan, code, and execute tasks.

iflytek/astron-agent

这是一个 AI 智能体 方向的 Java 开源项目，主要用于：Enterprise-grade, commercial-friendly agentic workflow platform for building next-generation SuperAgents

Enterprise-grade, commercial-friendly agentic workflow platform for building next-generation SuperAgents

Maciek-roboblog/Claude-Code-Usage-Monitor

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Real-time Claude Code usage monitor with predictions and warnings

Real-time Claude Code usage monitor with predictions and warnings

MiroMindAI/MiroThinker

面向复杂研究和预测任务的 Deep Research Agent，强调多步推理、信息检索与研究任务评估，适合关注通用研究智能体和 benchmark 的开发者。

MiroThinker is a deep research agent optimized for complex research and prediction tasks. Our latest models, MiroThinker-1.7, achieves 74.0 and 75.3 on the BrowseComp and BrowseComp Zh, respectively.

zilliztech/deep-searcher

开源深度研究 Agent，支持对私有数据进行推理与检索，可作为 Deep Research 的本地替代方案。

Open Source Deep Research Alternative to Reason and Search on Private Data.

withastro/flue

Astro 团队推出的沙箱型 Agent 框架，面向安全执行、工具调用和开发者工作流实验，适合观察前沿 agent runtime 设计。

The sandbox agent framework.

FareedKhan-dev/kimi-k3-in-c

这是一个 AI 智能体 方向的 C 开源项目，主要用于：A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.

A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.

2FastLabs/agent-squad

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Flexible and powerful framework for managing multiple AI agents and handling complex conversations

Flexible and powerful framework for managing multiple AI agents and handling complex conversations

steel-dev/steel-browser

为 AI Agent 和应用提供开源浏览器 API 与浏览器沙箱，适合网页操作、浏览器自动化和可控环境中的 browser agent 场景。

Open Source Browser API for AI Agents & Apps. Steel Browser is a batteries-included browser sandbox.

GoogleCloudPlatform/kubectl-ai

这是一个 AI 智能体 方向的 Go 开源项目，主要用于：AI powered Kubernetes Assistant

AI powered Kubernetes Assistant

ChatLab/ChatLab

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：Local-first chat history analyzer with AI. | 本地优先的 AI 聊天记录分析工具

Local-first chat history analyzer with AI. | 本地优先的 AI 聊天记录分析工具

deeplethe/utopia

这是一个 AI 智能体 方向的 Rust 开源项目，主要用于：World's first open-source enterprise world model.

World's first open-source enterprise world model.

yjh051108/dsh-routing-suite

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：dsh-routing-suite — injector + router-standard kit: install the runtime injector first, then the task-aware reasoning-mode router preset (measured P1-P23).

dsh-routing-suite — injector + router-standard kit: install the runtime injector first, then the task-aware reasoning-mode router preset (measured P1-P23).

genspark-ai/genoffice

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：Free, open-source AI Office suite: Docs, Sheets, Slides, PDF, Markdown and HTML editors with a built-in AI agent. Edits real .docx/.xlsx/.pptx files, converts PDF to Word on-device…

Free, open-source AI Office suite: Docs, Sheets, Slides, PDF, Markdown and HTML editors with a built-in AI agent. Edits real .docx/.xlsx/.pptx files, converts PDF to Word on-device…

microsoft/agent-governance-toolkit

微软 AI Agent 治理工具包，覆盖策略执行、零信任身份、沙箱和可靠性工程，适合评估与加固自主智能体的安全治理能力。

AI Agent Governance Toolkit — Policy enforcement, zero-trust identity, execution sandboxing, and reliability engineering for autonomous AI agents. Covers 10/10 OWASP Agentic Top 10.

microsoft/Webwright

微软开源的浏览器 Agent 框架，面向长链路 Web 任务与 SWE 风格自动化评测，适合研究复杂网页操作、浏览器规划和任务执行能力。

A simple SWE style browser agent framework that achieves SOTA results on long horizon web tasks.

Webwright

微软开源的浏览器智能体框架，面向长程网页任务、自动化操作和 SWE 风格评测场景。

A simple SWE style browser agent framework that achieves SOTA results on long horizon web tasks.

agent-infra/sandbox

面向 AI Agent 的一体化沙箱，把浏览器、Shell、文件、MCP 和 VS Code Server 集成到单个 Docker 环境中。

All-in-One Sandbox for AI Agents that combines Browser, Shell, File, MCP and VSCode Server in a single Docker container.

dataelement/dsh-desktop

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：DSHDesktop：DeepSeek Harness Desktop / DeepSeek Harness 桌面版.

DSHDesktop：DeepSeek Harness Desktop / DeepSeek Harness 桌面版.

coze-dev/coze-loop

Next-generation AI Agent Optimization Platform: Cozeloop addresses challenges in AI agent development by providing full-lifecycle management capabilities from development, debugging, and evaluation to monitoring.

iflytek/astron-rpa

这是一个 AI 智能体 方向的 Java 开源项目，主要用于：Agent-ready RPA suite with out-of-the-box automation tools. Built for individuals and enterprises

Agent-ready RPA suite with out-of-the-box automation tools. Built for individuals and enterprises

s1dashu/ip-as-logo-skill

这是一个 AI 智能体 方向的 开源 开源项目，主要用于：A compact Agent Skill for highly simplified, rounded, subtly neo-skeuomorphic IP mascot logos.

A compact Agent Skill for highly simplified, rounded, subtly neo-skeuomorphic IP mascot logos.

dtyq/magic

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：Magicrew. The first open-source all-in-one AI productivity platform (Generalist AI Agent + Workflow Engine + IM + Online collaborative office system)

Magicrew. The first open-source all-in-one AI productivity platform (Generalist AI Agent + Workflow Engine + IM + Online collaborative office system)

CopilotKit/OpenBot

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.

Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.

agentgateway/agentgateway

面向 AI Agent 和 MCP Server 的下一代代理网关，用 Rust 构建，适合研究智能体流量治理、API Gateway、反向代理与 Kubernetes 场景。

Next Generation Agentic Proxy for AI Agents and MCP servers.

ruc-datalab/DeepAnalyze

DeepAnalyze is the first agentic LLM for autonomous data science. 🎈你的AI数据分析师，自动分析大量数据，一键生成专业分析报告！

embabel/embabel-agent

面向 JVM 的 AI Agent 框架，支持 Java/Kotlin/Spring 生态、多智能体编排和生成式 AI 应用开发，适合企业团队在现有 JVM 技术栈内落地 Agent。

Agent framework for the JVM. Pronounced Em-BAY-bel /ɛmˈbeɪbəl/

gptme/gptme

终端里的持久化 AI Agent，可使用本地工具写代码、运行命令与浏览网页，适合开发者构建个人自动化助手和可扩展智能体。

Your agent in your terminal, equipped with local tools: writes code, uses the terminal, browses the web. Make your own persistent autonomous agent on top!

gptme

运行在终端里的个人 AI Agent，可调用本地工具、写代码、执行命令并浏览网页。

Your agent in your terminal, equipped with local tools: writes code, uses the terminal, browses the web. Make your own persistent autonomous agent on top!.

Archestra

面向企业 AI 的 MCP 注册表、网关与编排平台，强调工具接入、安全护栏和集中治理。

Enterprise AI Platform with guardrails, MCP registry, gateway & orchestrator.

XiaoDuoYa/codex-with-chatgpt

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：ChatGPT thinks. Codex works. Use ChatGPT as the planning brain while keeping the Codex harness.

ChatGPT thinks. Codex works. Use ChatGPT as the planning brain while keeping the Codex harness.

langroid/langroid

面向多智能体编程的 Python 框架，强调 Agent 之间的消息协作、RAG 与工具调用，适合构建复杂对话流和任务分解系统。

Harness LLMs with Multi-Agent Programming

Langroid

用于多智能体编程的 LLM 应用框架，帮助开发者组合代理、消息流和工具调用。

Harness LLMs with Multi-Agent Programming.

evalstate/fast-agent

面向构建、评估与编排智能体的 Python 框架，内置 MCP/ACP 支持，适合把工具调用、模型评测和多步骤工作流组合成可复用 Agent。

Code, Build and Evaluate agents - excellent Model and Skills/MCP/ACP Support

LazyAGI/LazyLLM

面向多智能体与 LLM 应用的 Python 框架，覆盖工具调用、RAG 和 Agent 编排，适合快速搭建原型和教学型多 Agent 工作流。

Easiest and laziest way for building multi-agent LLMs applications.

TracecatHQ/tracecat

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Open-source security automation platform for teams and AI agents

Open-source security automation platform for teams and AI agents

yvgude/lean-ctx

本地优先的 Agent 上下文控制层，决定 AI 能读取、记住和修改哪些内容，并提供 MCP 工具与审计凭据，适合降低 token 消耗和增强代码代理安全边界。

Control what your AI can see. LeanCTX (Lean Context) is the context intelligence layer for AI agents — one local Rust binary that decides what they read, remembers what they learn, guards what they touch, and proves what they save. 60–90% fewer tokens as the receipt. 76 MCP tools, 30+ agents, local-first.

yaojingang/GEOFlow

这是一个 AI 智能体 方向的 PHP 开源项目，主要用于：Open-source GEO content engineering and multi-site distribution system with AI tasks, RAG/semantic chunking, analytics, GEOFlow Agent and WordPress target publishing

Open-source GEO content engineering and multi-site distribution system with AI tasks, RAG/semantic chunking, analytics, GEOFlow Agent and WordPress target publishing

SkyworkAI/DeepResearchAgent

层级式多智能体深度研究系统，可自动拆解研究问题、多轮检索与推理，生成结构化研究报告。

DeepResearchAgent is a hierarchical multi-agent system designed for deep research tasks.

Ar9av/obsidian-wiki

让 AI Agent 围绕 Obsidian Wiki 构建和维护个人知识库的工具，适合将项目文档、上下文和长期记忆整理成可检索的数字大脑。

Framework for AI agents to build and maintain a digital brain through Obsidian wiki using Karpathy's LLM Wiki pattern

langchain-ai/langgraphjs

LangChain 官方的 TypeScript 图式智能体框架，用有状态图建模多步骤 LLM 应用，适合需要可恢复、可观测工作流的 Agent 项目。

Framework to build resilient language agents as graphs.

TanStack/ai

TanStack 生态的类型安全 AI SDK，支持流式聊天、工具调用、智能体和多模态应用，适合前端/全栈团队构建跨模型 AI 功能。

Type-safe, provider-agnostic TypeScript AI SDK for streaming chat, tool calling, agents, and multimodal apps across OpenAI, Anthropic, Gemini, React, Vue, Svelte, and Solid.

yilewang/llm-for-zotero

扎根 Zotero 文献库的研究 Agent 系统，把论文管理、RAG、文献分析和 MCP 工作流结合起来，适合学术阅读、资料整理和研究型知识助理。

A research agent system deeply rooted in your Zotero library.

qingchencloud/clawpanel

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：🦞 OpenClaw & Hermes Agent 多引擎 AI 管理面板 — 内置 AI 助手（工具调用 + 图片识别 + 多模态），一键安装 | Tauri v2 跨平台桌面应用 | 11 种语言

🦞 OpenClaw & Hermes Agent 多引擎 AI 管理面板 — 内置 AI 助手（工具调用 + 图片识别 + 多模态），一键安装 | Tauri v2 跨平台桌面应用 | 11 种语言

anthropics/commerce-agents

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：Reference blueprint for building shopping and merchant agents with Claude. Examples in retail, commerce, telecom, and entertainment included.

Reference blueprint for building shopping and merchant agents with Claude. Examples in retail, commerce, telecom, and entertainment included.

Player-YN/PawWork_ZhuaZhua

这是一个 AI 智能体 方向的 JavaScript 开源项目，主要用于：Paw Work - selection-first web agent for Chrome: select on the live page, describe the outcome, take away an editable office file. BYOK, sandboxed, no server.

Paw Work - selection-first web agent for Chrome: select on the live page, describe the outcome, take away an editable office file. BYOK, sandboxed, no server.

Nanako0129/sepia

这是一个 AI 智能体 方向的 Python 开源项目，主要用于：De-AI writing skill for any Agent Skills-compatible agent (77+ via the Skills CLI), with native plugins for Claude Code, Codex, Grok Build, and Antigravity. Narrative-architecture…

De-AI writing skill for any Agent Skills-compatible agent (77+ via the Skills CLI), with native plugins for Claude Code, Codex, Grok Build, and Antigravity. Narrative-architecture…

alvinreal/awesome-autoresearch

围绕 autonomous improvement loops、research agents 和自我改进研究系统的精选清单，适合快速追踪 Autoresearch、科学发现与研究智能体方向的开源资源。

A curated list of autonomous improvement loops, research agents, and autoresearch-style systems inspired by Karpathy's autoresearch.

samanhappy/mcphub

统一管理和动态编排多个 MCP servers/API 的 Hub，可拆分端点并配置路由策略，适合需要集中管理 MCP 工具链的场景。

A unified hub for centrally managing and dynamically orchestrating multiple MCP servers/APIs into separate endpoints with flexible routing strategies.

MCPJam/inspector

面向 MCP server、MCP 应用和 ChatGPT apps 的测试评估与调试平台，可直接对话、检查协议调用并排查 MCP 集成问题。

Testing and evaluation platform to chat, inspect, and debug MCP servers, MCP apps, and ChatGPT apps.

GCWing/BitFun

桌面级 Agent runtime 与应用套件，内置 Code Agent、Cowork Agent 和 Computer Use 能力，适合研究本地桌面 Agent 运行时。

BitFun is a desktop-grade Agent runtime and a ready-to-use suite of desktop Agent applications with built-in Code Agent, Cowork Agent and Computer Use.

Notte

用于构建 Web agents 并部署 serverless 浏览器自动化函数的 Python 框架，适合可靠网页任务执行。

🌸 Best framework to build web agents, and deploy serverless web automation functions on reliable browser infra.

openakita/openakita

开源 AI 助手框架，提供 skills 与 agent architecture，适合搭建可扩展的个人/团队智能体工作流。

An open-source AI assistant framework with skills and agent architecture

yomorun/yomo

Serverless AI Agent Framework，结合 QUIC、MCP、函数调用和边缘分布式基础设施，适合构建低延迟、多地域的智能体运行与流式处理系统。

🦖 Serverless AI Agent Framework with Geo-distributed Edge AI Infra.

trpc-group/trpc-agent-go

Go 语言生产级 Agent 框架，覆盖 graph workflows、工具、记忆、A2A、AG-UI、MCP、评测与可观测性，适合后端团队搭建可运维智能体系统。

A Go framework for building production agent systems with graph workflows, tools, memory, A2A, AG-UI, MCP, evaluation, and observability.

open-mercato/open-mercato

这是一个 AI 智能体 方向的 TypeScript 开源项目，主要用于：AI-Engineering Foundation Framework built with AI and designed for AI. Hundreds of architectural and domain decisions (multi-tenancy, RBAC, event flow, pricing, sales pipeline,CRM

AI-Engineering Foundation Framework built with AI and designed for AI. Hundreds of architectural and domain decisions (multi-tenancy, RBAC, event flow, pricing, sales pipeline,CRM

openkursar/hello-halo

7×24 桌面 AI Agent，提供可视化助手、远程访问、文件管理和内置 AI 浏览器，适合桌面端个人智能体产品参考。

7×24 Desktop AI Agent for Everyone. Visual AI assistant with remote access, file management, and built-in AI browser.

OpenOSINT/OpenOSINT

AI 驱动的开源情报（OSINT）智能体，内置交互式 REPL、MCP Server 和 CLI，集成 16 种情报工具，支持 Claude、GPT-4 和本地模型，适合安全研究和威胁情报分析。

AI-powered OSINT agent with interactive REPL, MCP server, and CLI. 16 tools. Works with Claude, GPT-4, or local models.

inkeep/agents

可视化 No-Code Builder 与 TypeScript SDK 双向同步的 AI Agent 构建平台，适合快速交付客服助手、多智能体工作流和可自托管 Agent 原型。

Create AI Agents in a No-Code Visual Builder or TypeScript SDK with full 2-way sync. For shipping AI assistants and multi-agent AI workflows.

preset-io/agor

面向团队的智能体编排画布，可在多人界面中管理 Claude Code、Codex、Gemini 会话和 git worktree，适合可视化跟踪 agentic 开发过程。

Orchestrate Claude Code, Codex, and Gemini sessions on a multiplayer canvas. Manage git worktrees, track AI conversations, and visualize your team's agentic work in real-time.

dohooo/helmor

本地多智能体软件开发工作台，可协调 Claude Code、Codex 等编码代理参与项目任务，适合探索多代理协作、代码审查和开发流程编排。

Open-source local workbench for multi-agent software development.

gastownhall/gascity

用于多智能体编码工作流的编排构建 SDK，关注 agent team、任务流和开发自动化，适合观察 coding agent 协作层设计。

Orchestration-builder SDK for multi-agent coding workflows

wanxingai/LightAgent

轻量级 AI Agent 框架，支持记忆、MCP、技能和多智能体协作，适合快速搭建可扩展的自主任务处理与工具调用系统。

LightAgent: Lightweight AI agent framework with memory, mcp & skill. Supports multi-agent collaboration, self-learning, and major LLMs (OpenAI/DeepSeek/Qwen). Open-source with MCP/SSE protocol integration.

modu-ai/moai-adk

面向 Claude Code 的 SPEC-first Agentic Development Kit，内置多智能体、技能包和质量门禁，适合规范化自动开发流程。

SPEC-First Agentic Development Kit for Claude Code — 24 AI agents + 52 skills with TDD/DDD quality gates, 16-language projects, 4-language docs. Go CLI, zero deps.

vm0-ai/vm0

面向真实工作任务的 AI teammate 项目，强调可信协作和任务执行体验，适合参考智能体产品形态、工作台交互和自动化流程。

Zero, your trustworthy AI teammate for real work.

LiuMengxuan04/MiniCode

轻量级终端编码助手，提供类似 Claude Code 的工具循环、TUI 和任务执行体验，适合研究本地 coding agent 架构与命令行交互。

A lightweight terminal coding assistant with Claude Code-like workflow, tool loop, and TUI architecture.

fuyuxiang/echo-agent

面向私有部署的 Agent 操作系统，由 planner、coder、researcher、operator 等角色组成，支持认知记忆、安全审批、A2A 与 MCP。

Echo Agent 是一个面向私有部署的 Agent 操作系统。一支由 planner、coder、researcher、operator 组成的 AI 团队，部署在你自己的服务器上，7×24 常驻运行。认知级记忆系统（四层架构 + 遗忘曲线 + 矛盾检测）让它越用越懂你；LLM 驱动的安全审批让高风险操作摆脱死板规则；原生 A2A 和 MCP 协议让它成为 Agent 互联网中的一等公民。从微信到 Telegram，从 CLI 到 Gateway API，一个入口统一所有。

asklokesh/loki-mode

多智能体自治 SDLC 框架，可从需求说明、GitHub issue、OpenAPI 或一句话 brief 推进到应用交付，并提供质量门禁。

Multi-agent autonomous SDLC framework. Spec to deployed app. PRD, GitHub issue, OpenAPI/JSON/YAML, or one-line brief. 5 AI providers, 11 quality gates.

agents-flex/agents-flex

轻量级 Java AI 应用开发框架，帮助 Java 团队构建 LLM 应用与 Agent 能力，适合后端生态集成。

Agents-flex is A Lightweight Java AI Application Development Framework.

LocoreMind/locoagent

面向社交媒体运营的 AI 浏览器智能体，可在真实浏览器环境中自动执行发帖、互动和内容管理等操作，适合社媒自动化和智能运营场景。

AI-powered social media agent with real browser automation

ByteChef

开源 AI agent 编排与工作流自动化平台，在同一套流程里平衡自主执行和可控编排。

Open-source platform that unifies AI agent orchestration and workflow automation — autonomy and precision in one platform.

im4codes/imcodes

面向 AI Agent 的即时协作与共享上下文工具，支持记忆、监督执行和跨提供商审计，适合多 Agent/多 CLI 协同。

The IM for agents. Shared Agent Context & Memory, supervised execution, and cross-agent audit across AI providers.

zamalali/DeepGit

面向 GitHub 仓库发现的 Deep Research Agent，结合搜索、LangGraph/LangChain 与 LLM 推理，适合做开源项目调研和技术选型线索挖掘。

Deep research agent to help you find the best GitHub repositories 🕵️!

agentic-community/mcp-gateway-registry

企业级 MCP Gateway 与 Registry，集中管理 AI 开发工具、OAuth 鉴权、动态工具发现和审计式访问，适合治理多 MCP 服务。

Enterprise-ready MCP Gateway & Registry that centralizes AI development tools with secure OAuth authentication, dynamic tool discovery, and unified access for both autonomous AI agents and AI coding assistants. Transform scattered MCP server chaos into governed, auditable tool access with Keycloak/Entra integration.

hyper-mcp-rs/hyper-mcp

Rust 编写的高性能 MCP Server，通过 WebAssembly 插件扩展能力，适合需要隔离执行和可插拔工具的 Agent 平台。

📦️ A fast, secure MCP server that extends its capabilities through WebAssembly plugins.

Purewhiter/mobilegym

面向移动端 GUI Agent 研究的高并行仿真平台，可在浏览器中运行安卓模拟器，提供可验证的评测环境，适合移动智能体训练、评测和 sim-to-real 迁移研究。

MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research

LvcidPsyche/auto-browser

MCP 原生的开源浏览器智能体，让 AI Agent 使用真实浏览器并保留 human-in-the-loop 控制，适合网页自动化、Playwright、远程浏览器和自托管 Agent 场景。

Give your AI agent a real browser — with a human in the loop. Open-source MCP-native browser agent.

unbrowse-ai/unbrowse

面向浏览器智能体的 MCP/SDK 工具，可从浏览器流量中发现 API 并生成可调用技能，适合把网页操作转成更稳定的自动化接口。

Unbrowse — api native browser MCP/SDK for any agent. Auto-discovers APIs from browser traffic, generates skills on the fly to call APIs directly 100x faster, 80% cheaper locally.

browser-use/desktop

Browser Use 的桌面应用形态，把浏览器智能体操作封装成可直接运行的本地应用，适合体验和验证网页任务自动化流程。

Browser Use Desktop App

chubbyguan/chubbyskills

面向中文全渠道内容采集的 13 个 AI Skill，支持抖音、B站、小红书、公众号、X 和播客等内容源，可将图文、视频、字幕采集进个人知识库，附带 MCP Server。

13 AI Skills for ingesting Chinese content from Douyin, Bilibili, Xiaohongshu, WeChat, X, and podcasts into personal knowledge base with MCP server.

OpenManus

受 Manus 类任务自动化启发的开源通用 agent 项目。

Open-source general agent project inspired by Manus-style task automation.

ruvnet/agent-harness-generator

AI Agent 的元脚手架工具，可通过 npx 快速生成带独立 CLI、MCP Server、记忆层和学习循环的专属 Agent Harness，适合快速搭建定制化智能体框架。

Meta-harness for AI agents — scaffold your own focused, branded agent harness with its own npx CLI, MCP server, memory, learning loop, and witness layer.

framerslab/agentos

TypeScript AI Agent 框架，内置认知记忆、运行时工具生成、多智能体编排和多家 LLM Provider 接入，适合研究更完整的 Agent 操作系统形态。

TypeScript AI agent framework: cognitive memory, runtime tool forging, multi-agent orchestration, 11 LLM providers.

sandiiarov/skill-creator

运行时工具生成器，可将任意 MCP Server、OpenAPI 规范或 GraphQL 端点即时转化为可用 CLI，适合快速扩展 Agent 工具链和协议适配。

Turn any MCP server, OpenAPI spec, or GraphQL endpoint into a CLI at runtime.

maze-agent/Maze

分布式 LLM Agent 框架，面向多智能体协作、分布式执行和智能体系统实验，适合研究与工程原型。

A distributed framework for LLM agents

tt-a1i/hive

浏览器原生的 CLI 编码智能体协作层，可让 Claude Code、Codex、Gemini、OpenCode 等在同一任务中协同，适合多 Agent 开发流程。

Browser-native hive-mind for CLI coding agents — Claude Code, Codex, Gemini, and OpenCode collaborate as real PTY processes via a team protocol.

tinykit-studio/tinykit

Self-hosted Lovable/v0 alternative. Realtime database & storage included.

Cranot/roam-code

本地代码库智能 CLI 与 MCP Server，可生成 SQLite 代码图谱、暴露多语言代码检索和变更安全检查能力，适合给 Claude Code、Codex 等编码 Agent 补充项目上下文。

Local codebase intelligence CLI + MCP server for AI coding agents: SQLite code graph, 28 languages, 238 commands, 224 MCP tools, change-safety gates, audit evidence, zero API keys.

cordum-io/cordum

开源智能体控制平面，聚焦 Agent 执行前策略、审批和治理，适合需要把自主智能体接入企业规则与安全边界的团队。

The open agent control plane. Govern autonomous AI agents with pre-execution policy enforcement.

saltbo/agent-kanban

面向 AI workforce 的 Agent-first 看板，把 Claude Code、Codex、Gemini 等编码代理任务纳入任务管理与编排视图，适合多 Agent 开发协作。

An agent-first task board, Mission control for your AI workforce.

apache/flink-agents

Apache Flink 生态的 Agentic AI 框架，关注流式数据、事件驱动和企业级智能体编排，适合实时任务自动化探索。

Flink Agents is an Agentic AI framework based on Apache Flink

paperclipinc/openclaw-operator

用于部署和管理 OpenClaw AI agent 实例的 Kubernetes Operator，覆盖安全、可观测和生命周期管理，适合生产化 Agent 运维参考。

Kubernetes operator for deploying and managing OpenClaw AI agent instances with production-grade security, observability, and lifecycle management.

ucsandman/DashClaw

AI Agent 治理运行时，可拦截动作、执行安全策略、加入审批并生成审计轨迹，适合需要 guardrails 和合规记录的 Agent 工作流。

🛡️The governance runtime for AI agents. Intercept actions, enforce guard policies, require approvals, and produce audit-ready decision trails.

sachitrafa/YourMemory

带艾宾浩斯遗忘曲线衰减机制的 Agent 记忆系统，结合 MCP、pgvector 与长期记忆评测，适合研究智能体记忆保留和召回策略。

Agentic AI memory with Ebbinghaus forgetting curve decay. +16pp better recall than Mem0 on LoCoMo.

MigoXLab/webqa-agent

用于网站性能、功能和交互体验审计的自主浏览器 QA Agent，支持 GUI/CLI 测试评估，适合工程师和 vibe-coding 项目做自动化网页质量检查。

Autonomous web browser agent that audits performance, functionality & UX for engineers and vibe-coding creators. 网站自主评估测试 Agent，支持 GUI/CLI 一键完成性能、功能使用与交互体验的测试评估

step-security/dev-machine-guard

开发机安全扫描工具，可检查 AI Agents、MCP Servers、IDE 扩展和可疑包，适合引入智能体工具后的本地安全体检。

Scan your dev machine for AI agents, MCP servers, IDE extensions, and suspicious packages - in seconds.

cyanheads/mcp-ts-core

面向 MCP Server 的 TypeScript 核心框架，提供声明式工具定义、认证、多传输等能力，适合构建 Agent 可调用的工具服务。

Agent-native TypeScript framework for building MCP servers. Declarative definitions with auth, multi-backend storage, OpenTelemetry, and first-class support for Bun/Node/Cloudflare Workers.

Wide-Moat/open-computer-use

为任意 LLM 提供托管 Docker 工作区、浏览器、终端、代码执行和子 Agent 的 MCP Server，适合自托管 computer-use 场景。

MCP server that gives any LLM its own computer — managed Docker workspaces with live browser, terminal, code execution, document skills, and autonomous sub-agents. Self-hosted, open-source, pluggable into any model.