Projects

AI驱动的科研前沿探索平台，让更强大的AI推动科学边界，构建更强大的共享未来。支持自动化科研流程、论文分析和实验设计。

迷你CLI语义搜索引擎，专为本地文档、知识库和会议笔记设计。追踪当前最先进的检索方法，完全本地运行，无需云服务。17k+ Stars。

编码Agent的便携式记忆层工具。为自主编码Agent（如Claude Code、Codex等）提供跨会话持久化记忆，支持项目上下文管理、任务状态追踪和知识图谱构建。

AI驱动的TradingView MCP服务器，支持实时加密货币和股票筛选、技术指标分析、布林带智能分析和K线形态识别。兼容Claude Desktop等AI助手，支持币安、KuCoin、Bybit等多交易所。

支持100+语言的强大OCR工具包，一键将PDF和图片转为结构化数据，轻量高效，是连接文档与LLM的桥梁

基于 LLM 的多 Agent 协作软件开发框架，模拟完整软件公司流程，自动化从需求到代码的全过程

OpenAI Codex CLI 的工作流增强层，支持 hooks、agent teams、HUD 等功能扩展，让 Codex 更强大

On-device LLM inference engine by Google AI Edge team, built in C++ for efficient large language model execution on mobile and edge devices.

Blazing fast fuzzy file finder for AI agents and Neovim with built-in memory — reduces token usage by intelligently ranking files based on frecency, git status, and definitions. MCP protocol supported.

AI agent toolkit by the creator of libGDX: coding agent CLI, unified multi-provider LLM API, TUI & web UI libraries, Slack bot, and vLLM pod management — everything you need to build AI agents.

Google出品的轻量级端侧LLM推理引擎，C++实现，专为移动设备和嵌入式平台优化的大语言模型运行时

Google官方端侧AI模型体验平台，支持在手机上离线运行Gemma 4等开源大模型，完全本地推理、隐私安全、速度飞快

下一代全模态RAG框架，基于LightRAG构建，支持文本、图片、表格、公式、图表等多模态文档的统一处理与检索增强生成

全能AI Agent开发工具箱：编码Agent CLI、统一多LLM API、终端与Web UI库、Slack机器人、vLLM部署管理，一站式构建AI Agent应用

Imbue AI 开源的 Agent 管理 CLI 工具，用于在终端中管理和编排多个 AI Agent 的运行。轻量级设计，适合开发者快速调度 Agent 任务。

通义千问团队开源的终端AI编码Agent，类似 Claude Code 的开源替代品。支持在终端中进行代码生成、编辑、调试等操作，基于 Qwen 模型。

GitHub 官方推出的多平台 SDK，用于将 Copilot Agent 能力集成到任意应用和服务中。支持多种语言和平台，提供标准化的 AI 编码助手接入方式。

成为顶级AI/ML研究工程师的系统化学习路线，覆盖数学、计算机科学和AI核心知识体系，含交互式内容。

Every公司开源的复合工程插件，为Claude Code、Codex等AI编码工具提供团队协作增强，支持多Agent工程化工作流。

AI Agent技能：自动跨Reddit、X、YouTube、HN、Polymarket等平台研究任意话题，并生成综合摘要报告。18k+ stars。

LLM应用的上下文优化层，自动压缩和管理上下文窗口，减少Token消耗同时保持输出质量。支持多种压缩策略。

Every公司开源的复合工程插件，支持Claude Code和Codex，实现团队级多Agent协作开发工作流

为AI Agent打造的最快最精准文件搜索工具包，支持Neovim/Rust/C/NodeJS，让Agent高效定位代码

基于机器学习的漫画翻译器，用Rust编写，自动识别漫画文字气泡并翻译，支持多语言

LLM应用的上下文优化层，智能压缩和管理上下文窗口，减少token消耗同时保持输出质量

Vercel实验室出品的Bash for Agents工具，让AI Agent能够更安全高效地执行Bash命令，适合Agent自动化场景。

微软开源的AI Agent构建、编排与部署框架，支持Python和.NET，提供多Agent工作流支持，适合企业级Agent应用开发。

最佳Agent编排框架（原oh-my-opencode），支持多Agent协作、插件系统和丰富的开发者工具，48000+ stars的热门项目。

Claude Code / Codex / Cursor 等编码Agent的性能优化系统，包含技能树、本能反应、记忆管理、安全策略和研究优先开发方法论。GitHub本周增长23000+ stars。

开源AI Agent编排平台，提供可视化界面来构建、部署和管理AI Agent工作流，是你AI团队的中央智能层

基于屏幕活动的AI Agent框架，通过记录和理解你的屏幕操作来驱动自动化Agent，让AI真正知道你在做什么并为你工作

AI编码Agent Token用量追踪CLI工具，支持Claude Code、Codex、OpenClaw、Gemini、Cursor等主流AI编码助手的token消耗监控，提供全球排行榜和2D/3D贡献图

高性能 AI 数据引擎，支持图片、音频、视频和结构化数据的大规模处理，Rust 核心 + Python 接口

Apple MLX 框架的 LLM 推理工具，让你在 Mac 上本地高效运行各种大语言模型

面向 AI 的数据转换框架，支持增量处理，性能极高，适合构建 RAG、搜索和分析的数据管道

Vercel Labs 出品的 AI Agent 浏览器自动化 CLI 工具，让 AI 代理能够直接控制浏览器完成网页操作任务

开源AI编码Agent，支持终端和桌面应用，多语言界面，一键安装，兼容所有主流LLM提供商，是Claude Code和Cursor的开源替代。

4.8万星的开源Agent编排工具，支持多模型混用（Claude/Kimi/GLM/GPT/Gemini），自动任务分配和纪律执行，被称为比Claude Code更快的编码Agent。

港大EMNLP2025论文实现，轻量级高效RAG框架，基于知识图谱增强检索，支持图存储、重排序、多模态数据处理和OpenSearch后端。

微软官方多Agent框架，支持Python和.NET双语言，提供图编排工作流、流式处理、检查点、人机交互等企业级能力，已整合Semantic Kernel和AutoGen。

一个 Agentic 技能框架和软件开发方法论，帮助 AI 编码 Agent 获得可复用的超能力技能包

将整个代码仓库打包成一个 AI 友好的单文件，方便喂给 Claude/ChatGPT/Gemini 等大模型进行代码分析

构建、部署和编排 AI Agent 的可视化平台，支持拖拽式工作流设计，是 AI 团队的中央智能调度层

Google Research 开发的时间序列基础模型，预训练的通用时序预测模型，可零样本迁移到各种预测任务。

开源 AI 平台，支持对接所有主流 LLM，提供高级 AI 对话功能，内置文档检索、权限管理和企业级部署能力。

LLM 应用的上下文优化层，帮你智能管理和压缩 prompt 上下文，在有限 token 预算内最大化信息密度。

GitHub官方发布的Copilot Agent多平台SDK，可将Copilot能力集成到任意应用和服务中

最强Agent增强框架(4.8万星)，为各种编码Agent提供统一的插件系统、团队协作和HUD界面

Vercel Labs推出的Agent Bash工具，让AI Agent安全高效地执行Bash命令，为Agent提供标准化的终端交互能力

Imbue开源的AI Agent管理CLI工具，像git管理代码一样管理编码Agent，支持SSH/Docker/Modal等多环境部署

通义千问团队开源的终端AI编码Agent，类似Claude Code的开源替代品，支持本地运行和多模型切换

An open source, extensible AI agent by Block that goes beyond code suggestions — install, execute, edit, and test with any LLM

Microsoft official open-source framework for building and orchestrating AI agents and multi-agent workflows with Python and .NET support

Efficiently run and fine-tune Vision Language Models on Mac using Apple MLX framework, supporting multiple mainstream VLM architectures

AI Agent 技能插件，可自动调研 Reddit、X、YouTube、HN、Polymarket 等多平台内容，生成有依据的综合摘要报告。

LLM应用的上下文优化层，帮助开发者智能压缩和管理上下文窗口，减少token消耗同时保持输出质量。

收录了 ChatGPT、Claude、Gemini、Grok 等主流AI产品的系统提示词泄露合集，定期更新。是学习顶级AI产品提示词设计的宝藏资源。

系统化的 AI/ML 研究工程师学习路线，覆盖数学基础、计算机科学核心和 AI 前沿，适合从零成长为顶尖 AI 研究工程师。

全能 LLM 命令行工具，集成 Shell 助手、Chat-REPL、RAG、AI 工具和 Agent 功能，支持 OpenAI/Claude/Gemini/Ollama 等主流模型。

基于机器学习的漫画翻译工具，用 Rust 编写，支持自动检测气泡文字、OCR 识别和翻译渲染，专为漫画场景优化。

基于 Rust 构建的 AI 编程助手，支持 Claude、GPT、Gemini、DeepSeek 等 300+ 模型，在终端中实现智能结对编程。

Every 公司开发的复合工程插件，适用于 Claude Code、Codex 等 AI 编码工具，实现多 Agent 协作工程化。

AI Agent 记忆管理库，为构建有状态的 Agent 提供记忆持久化和个性化交互能力。

高精度 OCR 模型，擅长处理复杂表格、表单和手写文字，支持完整版式还原。

免费开源的屏幕录制与演示工具，Screen Studio 的开源替代品。无订阅、无水印，支持商业使用。

将文档网站、GitHub仓库和PDF自动转换为Claude AI技能包，支持冲突检测

一站式开源LLM微调、评测和部署平台，支持 Qwen3、DeepSeek-R1 等主流开源模型

你的24/7 AI员工团队 - 低代码多Agent AI平台，支持自动规划、研究、编码，可投递到 Telegram/Discord/WhatsApp

微软推出的 Agent 包管理器，为 AI Agent 生态提供标准化的技能包发现、安装和版本管理方案

本地化深度研究工具，SimpleQA基准达~95%准确率。支持Ollama等本地模型，搜索arXiv、PubMed等10+数据源，全程本地加密。

在Mac上使用Apple MLX框架运行和微调视觉语言模型(VLM)。支持主流开源VLM，本地推理无需GPU服务器。

为AI编程Agent设计的极速文件搜索工具，内置记忆功能减少token消耗。支持MCP协议，可集成Claude Code、Codex等。

微软推出的AI Agent包管理器，让你像管理npm包一样管理和分发AI Agent技能包。支持发现、安装、版本管理Agent工具。

社区维护的 Claude Code 技能、钩子、斜杠命令和插件精选列表，36k+ Stars

将 Cursor 免费 API 转换为 OpenAI/Anthropic 兼容格式的代理服务，支持 Claude Code 工具调用和图片

Roboflow 开源的计算机视觉工具库，提供检测、跟踪、标注等可复用组件，37k+ Stars

Datalab 出品的高精度 OCR 模型，能处理复杂表格、表单和手写体，支持完整版式还原，本周 2.4k Star 增长

成为顶尖 AI/ML 研究工程师的完整学习路线图，涵盖数学、计算机科学和人工智能的核心知识体系，2.8k Star

基于 CloudFlare 的免费临时域名邮箱，支持收发邮件、附件、IMAP/SMTP 和 Telegram Bot 集成，7.9k Star

Office 出品的 Claude Code/Codex 复合工程插件，支持多Agent团队协作、自动任务拆解和质量反馈循环，本周 1.4k+ Star

开源 AI 安全黑客工具，自动发现和修复应用漏洞，23k+ Star 的热门安全项目

为 AI Agent 构建有状态记忆的开源库，让 Agent 能记住上下文、用户偏好和历史交互

开源屏幕录制工具，可创建精美 Demo 视频，无水印、免费商用，Screen Studio 的开源替代品

OpenAI Codex CLI 的增强框架，支持 hooks、Agent 团队、HUD 面板等扩展功能，让 Codex 不再孤军奋战

Google Research开发的时间序列基础模型，专为时间序列预测设计，今日GitHub新增1100+ Star

微软开源的前沿语音AI模型，支持语音合成、语音识别等多种语音任务，今日GitHub新增1000+ Star

A collection of Codex skills shared by renowned iOS developer Dimillian, demonstrating how to build practical skill modules for AI coding assistants using Shell scripts.

A visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates for subagents, commands, skills, and orchestration workflows.

Microsoft open-source framework to train and optimize AI agents with zero code change. Supports all major agent frameworks with RL, automatic prompt optimization, and SFT algorithms.

Claude多Agent编排平台，支持智能Swarm部署、自主工作流协调和RAG集成，企业级架构。28K+ Stars本周热门。

自托管AI记账应用，利用LLM自动分析收据、发票和交易记录，支持自定义分类规则和多语言。3K+ Stars。

为AI Agent构建有状态记忆的开源库，支持长期记忆、用户画像和个性化对话管理，轻量级易集成。

自动化上传视频到社交媒体平台：抖音、小红书、视频号、TikTok、YouTube、Bilibili，支持定时发布和批量操作。

高精度OCR模型，专门处理复杂表格、表单和手写体，支持完整版式还原。由Surya/Marker作者出品，8K+ Stars。

高性能系统信息展示工具，neofetch 的现代替代品，用C语言编写，支持Linux/macOS/Windows等多平台

跨400+社交平台的用户名搜索工具，输入用户名即可追踪其在各大社交网络的账号

Claude Code 可视化实战指南——从基础概念到高级 Agent 编排，附带可复制的模板，15分钟快速上手

AI Agent 技能插件，能跨 Reddit、X、YouTube、HN、Polymarket 等平台研究任意话题近30天动态，生成有据可查的综合摘要。本周 GitHub 新增 10k+ star。

Letta AI 开发的 Claude Code 潜意识系统，为 Claude Code 添加持久记忆层，让它在多次会话间保持上下文连贯性。

AI 时代的记忆引擎，提供极速、可扩展的 Memory API。可以为任何 AI 应用添加长期记忆能力，支持语义搜索和上下文管理。

SakanaAI 推出的 AI 科学家 v2，能够通过智能体树搜索自动完成研讨会级别的科学发现。支持自动文献综述、假设生成、实验设计与执行、论文撰写全流程。

微软推出的开源前沿语音AI，支持高质量语音合成与语音理解，GitHub Trending热门项目

面向分析师、量化交易员和AI Agent的现代金融数据平台，支持多种数据源集成与分析

自托管AI虚拟伴侣容器，支持实时语音聊天、Minecraft和Factorio游戏陪玩，Web/macOS/Windows全平台支持，36k+ Star热门开源项目

test

AI Agent 可视化反馈工具，实时展示 Agent 的执行过程、决策路径和工具调用链，让 Agent 行为透明可观测。

Anthropic 官方发布的 Claude Agent SDK Python 版，提供构建 AI Agent 的标准化框架，支持工具调用、多轮对话和 Agent 编排。

开源 AI 安全黑客工具，自动发现并修复应用中的安全漏洞。利用 AI 模拟真实攻击者行为，进行自动化渗透测试。

新一代 OCR 模型，擅长处理复杂表格、表单和手写内容，完整保留文档布局结构

开源现代 CRM 平台，社区驱动的 Salesforce 替代方案，基于 TypeScript 全栈构建，已获 42k+ Star

开源AI对话平台，支持连接所有主流LLM，内置RAG知识库、文档解析、权限管理，适合企业级AI助手部署。

极速Whisper语音转文字工具，基于HuggingFace Transformers优化，速度比原版快数十倍，支持大文件批量转录。

实时换脸和一键视频深度伪造工具，只需一张照片即可实现实时人脸替换，支持多种AI模型和实时摄像头处理。

微软开源前沿语音AI框架，支持60分钟长音频一次转录（含说话人、时间戳）、实时TTS、50+语言，已集成HuggingFace Transformers

自主金融研究Agent，自动拆解复杂金融问题、调用实时市场数据、自我验证结果，像Claude Code但专为金融研究打造

AI自动科研系统v2：无需人工模板，自主生成假设、跑实验、写论文，首篇AI全自主论文已被ICLR Workshop接收

Chrome 官方出品的 DevTools MCP 服务器，让 AI 编程助手直接操控浏览器开发者工具，实现自动化调试

基于多智能体 LLM 的中文金融交易框架，TradingAgents 中文增强版，支持A股市场分析和交易策略

给 Claude Code 添加「潜意识」记忆层，让 AI 编程助手在会话间保持上下文记忆，由 Letta AI 团队开发

Anthropic 官方维护的 Claude Code 高质量插件目录，提供经过审核的插件集合，是学习插件开发的最佳参考

vLLM 多模态推理框架，支持文本、图像、音频等全模态模型的高效推理服务，是 vLLM 的下一代进化版本

可视化 Agent 构建与运行平台，支持可观测、可理解、可信赖的智能体开发，降低 Agent 开发门槛

高精度 OCR 模型，擅长处理复杂表格、表单和手写体，支持完整版面还原，文档数字化利器

Claude Code 多Agent协作编排框架，支持团队级任务分配与并行执行，让多个 Claude Code 实例协同完成复杂项目

test

利用AI大模型一键生成高清短视频的开源工具。支持自动生成视频脚本、配音、字幕和背景音乐，适用于批量内容创作场景。53000+ Star的热门项目，Python开发。

AI就绪的PDF解析器，自动化PDF数据提取和可访问性处理。开源Java项目，专注于将PDF转换为结构化的AI训练数据，支持表格、图表和复杂布局的智能解析。本周GitHub热门，9400+ Star。

Create and share 3D architectural projects. An open-source editor built with TypeScript for collaborative design.

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary.

The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows with enterprise-grade architecture, distributed swarm intelligence, RAG integration, and native Claude Code / Codex integration.

Anthropic 官方维护的高质量 Claude Code 插件目录，提供经过审核的插件集合，扩展 Claude Code 的能力边界。

LangChain 官方出品的 Agent 框架，内置规划工具、文件系统后端和子Agent生成能力，适合处理复杂的多步骤任务。

Claude Code 实时监控插件，可视化展示上下文使用量、活跃工具、子Agent状态和任务进度，开发者必备。

离线生存AI计算机，集成关键工具、知识库和AI能力，断网环境下也能提供信息支持和决策辅助。本周GitHub狂涨12000+ star。

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web, then synthesizes a grounded summary

Open-source online code editor with real-time collaboration

Memory engine and app that is extremely fast and scalable. The Memory API for the AI era.

面向物理空间的智能体操作系统（Agentic OS）。支持用自然语言控制人形机器人、四足机器人、无人机等硬件平台，构建与物理传感器（摄像头、激光雷达、执行器）无缝协作的多Agent系统。

基于 Next.js 的 AI 绘图应用，将 AI 能力与 draw.io 图表深度集成。支持通过自然语言命令创建、修改和增强各类图表，让架构图和流程图的绘制效率倍增。

零服务器代码智能引擎——完全在浏览器客户端运行的知识图谱创建工具。只需拖入 GitHub 仓库或 ZIP 文件，即可生成交互式代码知识图谱，内置 Graph RAG Agent 用于代码探索。

MiroMindAI 推出的深度研究Agent，专为复杂研究和预测任务优化。最新模型 MiroThinker-H1 在 BrowseComp 基准测试上达到 88.2 分，具备强大的网络浏览和信息综合能力。

A MCP server for Claude Code, Cursor, Windsurf to build n8n workflows from natural language.

Train a 26M-parameter GPT from scratch in just 2 hours with full pretraining, SFT, and RLHF pipeline.

An open-source SuperAgent harness by ByteDance that researches, codes, and creates with sandboxes, memories, tools and subagents.

统一的开源模型训练和推理 Web UI，支持 Qwen、DeepSeek、Gemma 等主流开源模型的本地微调和运行

从零到一构建一个类 Claude Code 的 Agent 编码工具，帮你理解 AI Agent 的核心原理：Bash is all you need

AI 编程助手性能优化系统，集成了 Skills、Instincts、Memory、Security 等模块，适用于 Claude Code、Codex、Cursor 等主流 AI 编程工具

Multi-agent LLM financial trading framework simulating real trading firm workflows with analysts, researchers and traders

Self-contained offline survival computer packed with critical tools, knowledge, and AI for anytime, anywhere use

Fully autonomous AI agent system for complex penetration testing tasks with multi-agent collaboration

Andrew Ng（吴恩达）团队的开源项目，专注于AI Agent的上下文管理和组织，帮助开发者更好地构建和管理Agent的知识与工具。

字节跳动开源的AI Agent上下文数据库，通过文件系统范式统一管理Agent所需的记忆、资源和技能，支持层级化上下文传递和自我进化。

LangChain官方出品的Agent框架，基于LangGraph构建，内置规划工具、文件系统后端和子Agent生成能力，适合处理复杂的多步Agent任务。

一个Agent技能框架和软件开发方法论，帮助构建具有专业能力的AI Agent。已获超10万Star，是目前最火的Agent Skills开源项目之一。

自动化在线赚钱流程的工具集，集成内容生成、发布和变现的完整工作流

vLLM团队推出的全模态模型推理框架，支持文本、图像、音频等多模态输入输出的高效推理

离线生存计算机——集成本地AI、关键知识库和实用工具，断网环境下也能提供信息检索和AI辅助

吴恩达团队出品的上下文管理中心，帮助 AI Agent 高效管理和检索上下文信息

LangChain 推出的深度 Agent 框架，内置规划工具、文件系统后端和子Agent生成能力，处理复杂任务

专为 AI 和自动化设计的无头浏览器，用 Zig 编写，性能极高，适合大规模网页抓取和自动化任务

LangChain 出品的开源异步编程 Agent，支持自主规划和代码修改，可作为 SWE-bench 替代方案

多Agent LLM金融交易框架，模拟真实交易公司的多角色协作进行投资决策

自托管AI记账应用，使用LLM分析收据、发票和交易，支持自定义提示词和分类

开源GPU加速物理模拟引擎，基于NVIDIA Warp构建，专为机器人研究和仿真研究人员设计

Andrew Ng 团队出品的上下文管理工具，帮助 AI Agent 高效管理和检索项目上下文信息

LangChain 开源异步编程 Agent，可自主完成代码修改、Bug修复等软件工程任务

Agent 技能框架与软件开发方法论，为 Claude Code 等 AI 编程工具提供可复用的技能包和最佳实践

完整的 AI 代理团队框架 - 涵盖前端开发、社区运营、创意注入等多种专业角色，每个 Agent 都有独立人格、工作流程和可交付成果

A Rust-written tool that can accurately recreate any real-world geographic location in Minecraft. Based on OpenStreetMap data, 10k+ Stars.

A lightweight meta-prompt and context engineering framework for Claude Code, a specification-driven development system. 35k+ Stars, a phenomenal project in AI programming.

An AI-ready PDF parser for automated PDF accessibility processing. Supports intelligent table recognition, multilingual OCR, and structured data output. Open source and free, currently #1 on GitHub trending.

An Agent coding framework created by the LangChain team, featuring built-in planning tools, filesystem backend, and sub-agent generation capabilities, suitable for handling complex multi-step Agent tasks. 15k+ stars.

An AI Agent context database open-sourced by ByteDance, unifying the management of Agent-required memories, resources, and skills through a filesystem paradigm, supporting hierarchical context and self-evolution. 15k+ stars.

A headless browser specifically designed for AI and automation, written in Zig for extreme performance. Suitable for Agent web interactions, data collection, and automated testing. 21k+ stars.

Microsoft's official 1-bit LLM inference framework, enabling efficient inference for ultra-low-bit quantized models, significantly reducing deployment costs and hardware requirements for large models. 35k+ stars.

A concise and versatile swarm intelligence prediction engine that predicts everything using collective wisdom. Based on Swarm Intelligence principles, supporting various prediction scenarios, with 34k+ Python stars.

An Agent memory system that can automatically learn and evolve from interactions. Open-sourced by Vectorize, helping AI Agents establish persistent, retrievable memory capabilities.

A concise and versatile swarm intelligence engine capable of predicting everything. Based on swarm intelligence algorithms, suitable for trend prediction and multi-agent collaborative decision-making scenarios.

A testing framework for Prompt/Agent/RAG, supporting red team testing and security scanning, capable of comparing performance across models like GPT, Claude, Gemini, with CLI and CI/CD integration support.

A headless browser specifically designed for AI and automation, written in Zig language, faster and lighter than traditional headless browsers, suitable for AI Agent web interaction scenarios.

An AI Agent context database open-sourced by Volcano Engine, unifying the management of agent-required memories, resources, and skills through a filesystem paradigm, supporting hierarchical context delivery and self-evolution.

Implement your favorite technologies from scratch! A curated collection of tutorials covering dozens of areas including 3D renderers, databases, compilers, operating systems, and more - the best practice for Feynman learning method.

A real-time status panel plugin for Claude Code, displaying context usage, tool activity, sub-agent status, and task progress without requiring additional windows.

An Agent framework by LangChain, ready out-of-the-box. Built-in planning, filesystem, shell, and sub-Agent capabilities, supporting MCP protocol extensions.

An LLM application evaluation and red team testing CLI tool, supporting automated prompt evaluation, model comparison, security vulnerability scanning, and CI/CD pipeline integration.

An AI Agent context database open-sourced by ByteDance's Volcano Engine, unifying the management of Agent memories, resources, and skills through a filesystem paradigm, supporting hierarchical context loading and visual retrieval.

An in-page GUI Agent open-sourced by Alibaba, controlling web interfaces with natural language without browser extensions or screenshots. Pure JS implementation, enabling quick integration of AI Copilot into products.

A swarm intelligence prediction engine based on multi-agent technology, simulating society and predicting future trends through high-fidelity digital worlds, supporting scenarios like public opinion analysis and financial forecasting.

An AI Agent context database open-sourced by ByteDance, unifying the management of Agent memories, resources, and skills through a filesystem paradigm, supporting hierarchical context transfer and self-evolution.

Automatically captures all operations in Claude Code programming sessions, compresses and summarizes them with AI, and injects them into future sessions, giving Claude cross-session memory capabilities. TypeScript implementation, 36k stars.

A simple and versatile swarm intelligence engine that predicts everything using Swarm Intelligence. Python implementation, supports various prediction scenarios, and is easy to extend. 3200+ GitHub stars today.

An LLM application evaluation and red team testing tool, supporting automated prompt testing, model comparison, security vulnerability scanning, and CI/CD integration, compatible with mainstream model providers like OpenAI, Anthropic, and Ollama.

A minimalist LLM training framework created by Karpathy, enabling the full workflow from tokenization to pre-training, fine-tuning, evaluation, and inference on a single GPU node. Train a GPT-2 level model for just $48 and interact with it via a ChatGPT-style Web UI.

An AI Agent context database open-sourced by ByteDance, using a filesystem paradigm to unify the management of Agent memories, resources, and skills, supporting hierarchical context loading and automatic session management, significantly reducing token consumption.

A swarm intelligence prediction engine based on multi-agent technology, simulating thousands of intelligent agents with independent personalities and memories in a high-fidelity parallel digital world to achieve accurate future trend predictions.

An AI Agent context database open-sourced by ByteDance, unifying the management of Agent memories, resources, and skills through a filesystem paradigm, supporting hierarchical context distribution and self-evolution.

A simple and versatile swarm intelligence engine capable of predicting anything. A prediction framework based on Swarm Intelligence, suitable for financial, weather, trend, and other prediction scenarios.

A headless browser designed for AI and automation, written in Zig, with extremely high performance. Ideal for building AI-driven web scraping, testing, and automation workflows.

An AI Agent context database open-sourced by ByteDance's Volcano Engine, unifying the management of Agent memories, resources, and skills through a filesystem paradigm.

Prompt/Agent/RAG testing and red team evaluation tool, supporting performance comparison and security scanning for mainstream models like GPT, Claude, Gemini, and Llama.

Microsoft's official 1-bit LLM inference framework, significantly reducing deployment costs and hardware requirements for large models, enabling efficient LLM operation on edge devices.

A simple and versatile swarm intelligence prediction engine, using Swarm Intelligence for predictive analysis of various events, with 17k+ GitHub stars this week.

A headless browser designed for AI and automation, written in Zig, extremely lightweight and high-performance, ideal for crawlers and Agent web interaction scenarios.

A one-stop RAG platform built on Langflow, Docling, and OpenSearch, offering full retrieval-augmented generation capabilities with a single package deployment.

An AI Agent context database open-sourced by ByteDance, unifying the management of Agent memories, resources, and skills through a filesystem paradigm, supporting hierarchical context distribution and self-evolution.

Unofficial Python API for Google NotebookLM, supporting batch importing source files, generating audio podcasts/videos/mind maps, and providing advanced features not exposed in the Web UI.

LLM application evaluation and red team testing tool. Supports automated prompt evaluation, model comparison, security vulnerability scanning, and integrates with CI/CD pipelines.

A self-evolving AI Agent developed by Nous Research, featuring built-in learning loops. Supports multiple platforms like Telegram/Discord/Slack, runs on $5 VPS, and is compatible with 200+ models.

From-scratch multi-agent public opinion analysis system covering 30+ global social platforms, automatically collecting and analyzing millions of comments. Features AI-driven monitoring, multimodal analysis, and Agent forum collaboration mechanisms.

Swarm intelligence prediction engine using multi-agent technology to simulate future scenarios via high-fidelity digital worlds. Upload seed materials, describe prediction needs in natural language, and receive detailed reports with interactive digital sandboxes.

Agent memory system enabling continuous learning and knowledge extraction from interactions, allowing AI Agents to reuse accumulated experience.

One-stop RAG platform built with Langflow, Docling, and OpenSearch, offering out-of-the-box retrieval-augmented generation solutions.

Microsoft's official 1-bit LLM inference framework, drastically reducing deployment costs and hardware requirements while enabling efficient large-model execution on CPUs.

Unofficial Python SDK for Google NotebookLM, supporting CLI and AI Agent integration for programmatic access to all NotebookLM features.

Official Agent framework for Alibaba's Qwen, supporting Function Calling, MCP protocol, code interpreter, RAG, and Chrome extensions.

Intelligent Agent framework by NousResearch that evolves through usage, supporting tool calls and multi-turn conversations.

Versatile swarm intelligence engine for universal prediction. Uses swarm algorithms for diverse prediction scenarios.

Swarm intelligence prediction engine using Swarm Intelligence for universal forecasting. Simple yet powerful, trending on GitHub with nearly 3000 stars today.

Agent memory system by Vectorize that continuously learns from interactions, giving AI Agents growing long-term memory.

Alibaba's open-source in-page GUI Agent for controlling web interfaces via natural language. Supports smart in-page operations, trending on GitHub with 1200+ stars today.

AI Agent by NousResearch that grows with you, featuring adaptive learning and skill expansion, starred 5k+ times.

Unofficial Python API for Google NotebookLM, enabling programmatic access to all features (including Web UI-exclusive ones) and integration with Claude Code, Codex, and other AI Agents.

Official Agent framework for Alibaba's Qwen, supporting Function Calling, MCP, code interpreter, RAG, Chrome extensions, etc. Built on Qwen 3.0+.

Multi-agent public opinion analysis tool for everyone, breaking information bubbles to reveal truth and predict trends. Built from scratch (no frameworks), starred 38k+ times.

Versatile swarm intelligence engine for universal prediction. Uses swarm algorithms for cross-domain predictive analysis, starred 16k+ times.

CLI tool for evaluating and red-teaming LLM applications. Automates prompt testing, model comparison, security vulnerability scans, and integrates with CI/CD. MIT-licensed, supports production apps with millions of users.

Alibaba's open-source in-page GUI Agent that controls web interfaces via natural language. No browser extensions or screenshots needed—pure JavaScript implementation adds AI copilot to products with just a few lines of code.

AI prediction engine using multi-agent technology to build high-fidelity parallel digital worlds where thousands of agents interact and evolve socially for accurate future projections. Supports financial signals, public opinion, novel endings, and more.

A self-evolving AI Agent by Nous Research, featuring a learning loop that creates skills from experience, auto-improves, retrieves cross-session memories, and models users. Runs on any platform/environment.

Agent development framework based on Qwen 3.0+, supporting Function Calling, MCP, Code Interpreter, RAG, Chrome extensions, etc. Serves as the backend for Alibaba's Qwen Chat.

Unofficial Python API for Google NotebookLM, enabling programmatic access to all NotebookLM features including podcast generation, batch imports, quiz exports, and other capabilities not exposed in the Web UI.

A versatile swarm intelligence engine that predicts everything through multi-agent simulation. Upload seed materials, describe prediction needs in natural language, and receive detailed reports with interactive high-fidelity digital worlds.

Andrej Karpathy's latest project: building the best ChatGPT alternative for just $100, featuring minimalist design and refined code.

An in-browser GUI Agent by Alibaba for controlling web interfaces with natural language, automating web operations without coding.

A concise and versatile swarm intelligence engine for predicting anything. Python implementation based on Swarm Intelligence theory, supporting diverse prediction scenarios. Trending on GitHub with 4400+ stars today.

A concise and versatile swarm intelligence engine for predicting anything. Open-source Python project based on collective wisdom methodologies.

An unofficial Python API and Agent skills for Google NotebookLM, enabling full programmatic access to NotebookLM features via Python, CLI, and AI Agents.

A real-time global intelligence dashboard—AI-driven news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface.

WiFi DensePose: Real-time human pose estimation, vital sign monitoring, and presence detection using ordinary WiFi signals—no cameras needed. Built with Rust for high performance and low latency.

An unofficial Python API for Google NotebookLM, enabling programmable access to NotebookLM's podcast generation, summarization, and other features.

A design language framework to make AI better at design, developed by Google engineers to help AI coding assistants generate more aesthetically pleasing UIs.

A concise and versatile swarm intelligence engine for predicting anything. Python implementation based on swarm intelligence algorithms, trending on GitHub with 2200+ stars today.

A collection of 169 production-ready skills and plugins for Claude Code / OpenAI Codex / OpenClaw, covering engineering, marketing, product, compliance, and more.

An unofficial Python API for Google NotebookLM, enabling programmatic notebook creation, source file uploads, and audio podcast summarization.

A real-time global intelligence dashboard—AI-driven news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface. Top GitHub trending this week with 33k+ stars.

An unofficial Python API for Google NotebookLM, enabling programmable access to NotebookLM's summarization, Q&A, and podcast audio generation features, suitable for batch processing and automation workflows.

An AI-native security testing platform integrating 100+ security tools, intelligent orchestration engines, and role-based testing capabilities. Built with Go, supporting full security testing lifecycle management.

A concise and versatile swarm intelligence engine for predicting anything. Based on swarm intelligence algorithms, supports stock market, weather, sports, and other prediction scenarios. Python implementation, trending on GitHub with 399 stars today.

A growing AI Agent framework by NousResearch, supporting autonomous learning and continuous evolution.

An adaptive web scraping framework supporting scenarios from single requests to large-scale crawling, with built-in anti-detection and intelligent parsing capabilities.

A web-based GUI Agent by Alibaba for controlling web interfaces with natural language. No browser extensions, screenshots, or multimodal models required—pure JavaScript implementation, integrable into any web app with one line of code. Ideal for building AI Copilots and smart form-filling.

A directory of Codex Agent Skills officially released by OpenAI, defining reusable instructions, scripts, and resource packages for AI Agents to automatically discover and execute tasks. Supports custom Skill creation and distribution.

A swarm intelligence prediction engine based on multi-agent technology, creating high-fidelity parallel digital worlds where thousands of Agents with independent personalities and long-term memories interact freely, simulating social evolution to accurately predict future trends. Supports financial forecasting, public opinion analysis, and narrative simulation.

Your AI second brain, a self-hosted personal AI assistant. Supports retrieving answers from webpages or documents, building custom Agents, and deep research.

A Chinese tutorial by Datawhale for building AI Agents from scratch, guiding you from principles to practice step-by-step.

A mobile GUI Agent series by Alibaba X-PLUG, supporting multimodal understanding of phone screens and automated operations, a benchmark project for GUI automation.

An adaptive AI Agent framework by NousResearch, supporting tool usage, memory management, and continuous learning, touted as an Agent that grows with you.

An enterprise-grade Prompt engineering framework by Microsoft, powering GitHub Copilot with constraint-driven AI workflows. Built-in RPI (Research→Plan→Implement) methodology, specialized Agents, and JSON Schema validation, suitable for individuals to large teams.

A Claude Code-based web novel writing assistant using RAG + entity graphs to address AI's memory and hallucination issues, supporting 2-million-word serials with built-in reader engagement tracking and visual Dashboard.

A Claude Code-based SEO long-form writing workstation with specialized commands like /research, /write, /optimize and 26 marketing Skills, integrating GA4 and Search Console data to mass-produce high-ranking blog content.

test

A self-hosted AI second brain supporting web search, local document Q&A, custom Agents, scheduled automation, and deep research, compatible with mainstream LLMs. ⭐ 33k

Huawei inclusionAI's open-source LLM reinforcement learning framework focused on enhancing reasoning and Agent capabilities, featuring speed and flexibility. ⭐ 4k

Datawhale's 'Building Agents from Scratch' tutorial systematically explains Agent principles and practices, ideal for beginners. Chinese tutorial with code examples. ⭐ 25.4k

Alibaba X-PLUG's mobile GUI Agent series supporting multimodal visual perception and autonomous smartphone operations, suitable for automated testing and mobile assistant scenarios. ⭐ 7.9k

An all-in-one security scanner for containers, K8s, code repositories, and cloud environments, detecting vulnerabilities, misconfigurations, and secret leaks. Essential for DevSecOps.

Ant Group's open-source LLM reasoning and Agent reinforcement learning framework for rapid training and enhanced model inference capabilities.

A Claude Code-based SEO long-form content workstation that automatically researches, writes, analyzes, and optimizes top-ranking blog content for any business scenario.

OpenAI's terminal programming Agent that writes and runs code via natural language in the command line

Anthropic's terminal-based AI programming tool for collaborative coding with Claude in the command line

An AI-first code editor deeply integrated with AI for smart coding and project understanding

A VS Code AI programming extension supporting multiple models and customizable workflows

An open-source AI search engine framework simulating human thinking for multi-step search and reasoning

Stanford's open-source AI long-form writing system that automatically researches and generates Wikipedia-level articles

Microsoft's next-gen AutoGen version with more flexible multi-Agent conversations and orchestration

A TypeScript AI Agent framework supporting workflows, RAG, and multi-Agent orchestration

Anthropic's open-source protocol standardizing connections between AI and external tools

An AI-powered web automation SDK open-sourced by Browserbase, controlling browsers with natural language

Enables AI Agents to browse the web like humans, performing clicks, inputs, and navigation

A lightweight Agent library open-sourced by HuggingFace, featuring concise APIs for building tool-calling Agents

A Pydantic-based Agent framework for type-safe AI application development

An AI-driven software development Agent capable of writing code, fixing bugs, and managing projects

A one-click tool for deploying LangChain chains and Agents as REST APIs

Extract clean text and metadata from unstructured documents such as PDF, Word, HTML and more

The open source RAG chatbot, powered by Weaviate, supports multiple embeddings and LLM

Open source personal AI assistant that can search notes, documents and the Internet, and supports self-hosting

Open source AI search engine, Perplexity replacement, supports multiple search modes

Enterprise-level AI search and Q&A assistant that connects internal company documents and tools

Cloud infrastructure that provides a secure sandbox environment for AI Agents to safely execute AI-generated code

Provides a platform for AI Agent to integrate 200+ tools, connecting GitHub, Slack, databases, etc.

An open source AI software engineering agent platform that can write and debug code like human developers

Quickly build a web demonstration interface for machine learning models, supporting multiple input and output types

Use Python to quickly build a web framework for data and AI applications, which can be deployed with just a few lines of code

Python framework to quickly build AI chat applications, supporting streaming, multi-modality and human-in-the-loop

LLM evaluation and red team testing tools to systematically evaluate prompt quality and model security

Add structured validation and safety guardrails to LLM output to prevent hallucinations and formatting errors

A structured data extraction library that allows LLM to reliably output JSON that conforms to the Pydantic model

AI-driven GitHub Bot automatically converts issues into PRs and handles bug fixes and small features

Run open source LLM’s desktop chat app on any device with complete offline privacy

Discover, download and run local LLM desktop applications on your laptop

Powerful self-hosted PDF tool that supports AI-assisted PDF editing and conversion

Self-hosted Google Photos alternative that supports AI face recognition and object search

Upload screenshots to automatically generate front-end code, supporting HTML/Tailwind/React/Vue, etc.

3Blue1Brown’s Python engine for producing mathematical animations can be used to generate teaching animations with AI assistance.

Human-level text-to-speech model with naturalness scores close to real people

Coqui's open source multi-language TTS model supports voice cloning and only requires 6 seconds of reference audio.

CogVideoX, the open-source Vincent video model of Zhipu, supports multiple resolutions and durations.

An open source project that reproduces Sora video generation, supporting video generation from text and images

Open source high-quality TTS toolkit, supporting multiple TTS engines and voices

Stability AI's latest Vincentian graph model is based on the MMDiT architecture and has greatly improved quality.

The third generation Llama model of Meta open source, version 8B and 70B, multi-language support

Microsoft's open source small language model, 3.8B parameter performance surpasses all models of the same level

Google's open source lightweight large model, version 2B and 9B, suitable for local deployment

In-depth exploration of open source MoE large models, 671B parameters and 37B activations, performance comparable to GPT-4

Mistral AI's open source, efficient large language model series achieves powerful performance with small parameters

Alibaba Tongyi Qianwen open source model, with multiple specifications to choose from and leading Chinese and English capabilities

Zero One Thousand Open Source's large language model series supports multiple specifications and multi-modalities.

Zhipu AI’s open source bilingual dialogue model supports 128K context and multi-modal understanding.

Kuaishou’s open source video generation model supports Wensheng videos and Tusheng videos.

Virtual human video generation framework based on diffusion model, supporting unlimited length videos

AI video generation tool, convert text and pictures into high-quality short videos

Black Forest Labs’ open source Vincentian graph model, quality comparable to Midjourney

Meta's universal image and video segmentation model, zero-shot segmentation of any object

A library of computer vision tools that simplify visualization and analysis of object detection, segmentation and tracking

Multilingual OCR and document analysis tools, supporting 90+ languages ​​with extremely high accuracy

Fast and accurate PDF to Markdown conversion tool that supports tables, formulas and code blocks

One-stop open source data extraction tool, convert PDF to Markdown and JSON, support OCR

IBM's open source document parsing tool converts PDF, DOCX, etc. to Markdown, supporting tables and formulas

Open source Zapier alternative that supports AI-driven automated workflow building

Workflow automation platform supports AI nodes and integrates LLM into automated processes

AI full-stack development tool produced by StackBlitz, build complete applications through conversations in the browser

The independent programming agent in VS Code can create files, execute commands, and use browsers

AI-driven full-stack application generator that generates complete deployable web applications through dialogue

Princeton’s open source automatic bug repair agent can automatically locate and fix code issues on GitHub

Open source AI software engineers, similar to Devin, can understand requirements and write code independently

Microsoft's open source code-first Agent framework is good at handling data analysis and automation tasks

Your second brain, use RAG to turn documents, web pages and notes into a conversational knowledge base

AI native data application development framework, supporting multi-model data-driven applications of AWEL and Agent

Enterprise-level knowledge base Q&A system based on large models, ready to use out of the box, supports multiple document formats

Enhanced version of ChatGPT clone, supporting multiple AI service providers, plug-ins and multi-modal interactions

An open source high-performance AI chat framework that supports multiple models, plug-in systems and knowledge bases

Professional-grade AI image generation workstation, providing canvas editing, ControlNet and workflow management

Stable Diffusion’s minimalist interface generates high-quality images without adjusting parameters.

Suno's open source text-to-audio model supports speech, music, background sound effects and non-verbal sounds.

Open source TTS model supports multi-language zero-sample speech cloning with latency as low as 150ms

Text-to-speech model optimized for dialogue scenarios, supports Chinese and English, and has natural rhythm

Wall-facing intelligent open source end-to-side large model series, 2B parameters are comparable to 7B effects, and supports mobile phone deployment

Asynchronous web crawler optimized for AI and LLM, automatically extracting structured data

Convert web pages into clean Markdown data available to LLM, supporting crawling of the entire website

Let LLM run code locally, control the computer through natural language, and execute Python, Shell, etc.

LLM inference engine implemented in pure C/C++, supports CPU and a variety of hardware, is lightweight and efficient

NVIDIA's LLM inference optimization library achieves ultimate inference performance on GPU

Microsoft's open source AI orchestration SDK integrates LLM with traditional programming languages ​​and supports plug-ins and plans.

LLM data connection framework connects private data with large models to build knowledge-enhanced AI applications

The AI ​​programming assistant in the terminal can directly edit code files through dialogue and supports Git integration.

Open source AI code assistant IDE plug-in, supports VS Code and JetBrains, and can be connected to any LLM

AI role-playing chat frontend for advanced users, supporting multiple LLM backends and rich customization options

Intelligent memory layer adds persistent personalized memory to AI applications, supporting user-level and session-level memory

End-to-end NLP/LLM application construction framework, supporting RAG, Q&A, semantic search, etc.

AI model routing platform, one API to access hundreds of AI models, automatically select the optimal model

Unified LLM API agent, calling 100+ LLMs in OpenAI format, including Azure, Anthropic, Ollama, etc.

Build an AI assistant framework with memory, knowledge and tools, supporting multi-modality and multi-Agent

Microsoft's open source multi-Agent dialogue framework supports automatic dialogue and collaboration between agents to solve problems.

A multi-agent collaboration framework that allows AI agents to work together like a team to complete complex tasks

Self-hosted AI code assistant, an open source alternative to GitHub Copilot, supports local deployment

Describe the requirements in one sentence, and AI automatically generates complete code projects to support iterative improvements.

AI-driven code editor, based on VSCode, integrates GPT-4 and Claude for intelligent code completion and dialogue

All-round AI document assistant, supports multiple LLM and vector databases, and provides a complete RAG pipeline

Visual framework for RAG and multi-Agent AI applications, supporting drag-and-drop construction and Python customization

Drag-and-drop LLM application building tool, visually orchestrate LangChain process, and build AI workflow with zero code

Open source ChatGPT desktop replacement, 100% offline operation, supports Llama, Mistral and other models

Open source OpenAI API alternative, supports local running of LLM, image generation and audio transcription, completely offline

100% privatized document question and answer system, no Internet required, supports local running of LLM for document interaction

The autonomous AI agent framework allows GPT-4 to complete complex tasks autonomously, automatically setting goals, decomposing steps and executing them.

NVIDIA's Edify 3D project can generate high-quality 3D models and complex scenes in 2 minutes, providing fine geometric meshes, clean topology, standardized UV maps, 4K textures and a complete PBR material system. Compared with other text to 3D tools, Edify 3D performs well in terms of shape and texture quality, efficiency and scalability, and supports generating 3D text or images.

DiffSplat is an AI tool for quickly generating 3D content. It can generate high-quality 3D content based on text or images within 1-2 seconds. It supports ControlNet for fine control and is very suitable for rapid iteration or real-time generated scenes.

TripoSG is a high-fidelity single image to 3D generation model with 1.5 billion parameters. It combines linear trajectory modeling of correction flow and Transformer architecture, which can stably model 3D object shapes and process photos, cartoons and sketches.

Tencent Hunyuan 3D generation model v2.5 is released. The number of parameters is increased from 1B to 10B, the number of effective patches is increased by 10 times, supports multi-view PBR veneers, 4K high-definition textures and fine-grained bump maps, optimizes the bone skinning system, and improves the efficiency of 3D animation generation.

Tencent has open sourced the HunyuanWorld-Mirror model, which supports the generation of high-quality 3D worlds from videos and multi-view images. It uses a feed-forward architecture to generate multiple three-dimensional representations at the same time, and is suitable for scenarios such as e-commerce and the digitization of museum cultural relics.

Alibaba launched LHM, a 3D digital human model that can be animated and reconstructed from a single image in a few seconds. It supports real-time rendering and posture control, with realistic effects and excellent detail processing.

GaussianCity is a tool for quickly generating ultra-large-scale 3D city models. The speed reaches 10.72 FPS, which is 60 times faster than existing solutions. It supports borderless city generation, uses BEV-Point as an intermediate representation, and combines with a space-aware decoder to achieve high-quality rendering.

StructLDM is an open source project that can generate a complete 3D human body model, supports the generation of different poses and perspectives, as well as the editing and modification of clothing and body shapes, allowing functions such as partial try-on and identity exchange.