1. moeru-ai/airi Not new today
💖🧸 Self hosted, you-owned Grok Companion, a container of souls of waifu, cyber livings to bring them into our worlds, wishing to achieve Neuro-sama's altitude. Capable of realtime voice chat, Minecraft, Factorio playing. Web / macOS / Windows supported.
First seen: February 28, 2026 | Streak: 8d
Analysis
Project AIRI is an open-source platform designed to create and host personalized "cyber-living" AI companions that can interact, chat, and play games like Minecraft and Factorio. Technically, the project leverages a sophisticated stack combining modern web technologies—such as WebGPU, WebAssembly, and Vue.js—with powerful native performance backends like NVIDIA CUDA and Apple Metal to handle complex AI tasks. This dual approach allows the software to run across web browsers, mobile devices, and desktop environments, providing a flexible framework for integrating large language models, speech recognition, and various avatar formats like Live2D and VRM.
This project is primarily aimed at AI enthusiasts, VTuber fans, and developers interested in building interactive digital humans who want a self-hosted alternative to closed-source systems like Neuro-sama. It is trending because it democratizes access to sophisticated digital companions, offering users full ownership and deep customization over their AI’s personality and capabilities. By providing a modular, cross-platform architecture that welcomes contributions from diverse skill sets, AIRI has cultivated a growing community focused on advancing the intersection of artificial intelligence and virtual embodiment.
Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.
First seen: March 07, 2026 | Streak: 1d
Analysis
Qwen-Agent is a comprehensive framework designed for developing Large Language Model (LLM) applications that leverage the instruction-following, planning, and memory capabilities of the Qwen model family. Technically, it provides a modular architecture with atomic components like `BaseChatModel` for LLM interaction and `BaseTool` for custom tool integration, enabling developers to build sophisticated agents capable of RAG, code execution, and MCP-compliant operations. The framework supports both cloud-based model services via DashScope and self-hosted deployments using OpenAI-compatible APIs like vLLM or Ollama, offering built-in support for Gradio-based GUI prototyping.
This project is highly beneficial for AI researchers, software developers, and enterprise engineers who need to build, evaluate, and deploy production-grade agentic workflows. It is trending because it provides a seamless bridge between cutting-edge open-source models like Qwen3.5 and practical, real-world utility through advanced tool-calling and multi-step reasoning capabilities. By offering a standardized way to integrate complex logic—such as document processing and web-based image manipulation—Qwen-Agent empowers users to quickly prototype and scale intelligent assistants tailored to specific domain requirements.
3. microsoft/hve-core Not new today
A refined collection of Hypervelocity Engineering components (instructions, prompts, agents, and skills) to start your project off right, or upgrade your existing projects to get the most out of all Copilots
First seen: March 05, 2026 | Streak: 1d
Analysis
HVE Core is an enterprise-grade framework designed to standardize and scale prompt engineering workflows within GitHub Copilot. It utilizes a structured "Research, Plan, Implement" (RPI) methodology to transition AI assistance from generic outputs to validated, constraint-based engineering artifacts. Technically, the project provides a comprehensive library of specialized agents, reusable prompts, and coding instructions that are governed by JSON schema validation and a CI/CD-integrated linting pipeline to ensure consistency and reliability.
This project is an essential resource for software development teams and organizations looking to integrate AI into professional engineering pipelines without sacrificing quality or security. It is trending because it addresses the "plausibility trap" of LLMs by implementing rigorous guardrails and specialized personas that allow developers to automate complex tasks confidently. By offering a modular, repeatable structure, HVE Core empowers teams to move beyond ad-hoc prompting and adopt a disciplined, high-velocity engineering approach that aligns with enterprise compliance standards.
CyberStrikeAI is an AI-native security testing platform built in Go. It integrates 100+ security tools, an intelligent orchestration engine, role-based testing with predefined security roles, a skills system with specialized testing skills, and comprehensive lifecycle management capabilities.
First seen: March 07, 2026 | Streak: 1d
Analysis
CyberStrikeAI is an AI-native security testing platform built in Go that orchestrates over 100 industry-standard security tools to automate the entire vulnerability lifecycle. By integrating natively with the Model Context Protocol (MCP) and supporting various LLM backends like GPT and DeepSeek, the platform transforms conversational commands into complex attack chains, knowledge retrieval, and actionable security reporting. It features a sophisticated modular architecture that includes role-based testing, YAML-based tool extensions, and a persistent SQLite backend to ensure that security workflows are both auditable and highly repeatable.
Security professionals, penetration testers, and CTF enthusiasts would benefit significantly from this project, as it effectively bridges the gap between manual security research and automated AI agent workflows. It is trending because it moves beyond simple chatbot interfaces by providing a structured, team-oriented environment that simplifies tool management, audit logging, and vulnerability tracking. By reducing the friction involved in chaining together disparate security utilities, CyberStrikeAI offers a powerful, collaborative framework that significantly enhances efficiency for modern security teams.
5. inclusionAI/AReaL Not new today
Lightning-Fast RL for LLM Reasoning and Agents. Made Simple & Flexible.
First seen: March 05, 2026 | Streak: 3d
Analysis
AReaL is an open-source, high-performance training system specifically designed for asynchronous reinforcement learning (RL) in large reasoning and agentic models. It achieves industry-leading speed and scalability by decoupling training components, allowing for seamless integration with various agentic runtimes and RL algorithms like GRPO, PPO, and RLOO. Technically, the system supports a wide range of hardware backends—including NVIDIA GPUs and Huawei Ascend NPUs—and offers both full-scale training architectures and a streamlined "AReaL-lite" version for rapid prototyping.
Researchers and developers looking to build and scale sophisticated AI agents will benefit significantly from AReaL’s flexible, "plug-and-play" API design. The project is currently trending due to its proven ability to produce state-of-the-art results in mathematical reasoning, coding, and autonomous search tasks, often outperforming or matching top-tier proprietary models. By providing a reproducible, affordable, and highly efficient framework, AReaL lowers the barrier to entry for those seeking to implement advanced self-evolving data synthesis and complex agentic workflows in their own AI projects.
基于 Claude Code 的长篇网文辅助创作系统，解决 AI 写作中的「遗忘」和「幻觉」问题，支持 200 万字量级 连载创作。
First seen: March 07, 2026 | Streak: 1d
Analysis
Webnovel Writer 是一个基于 Claude Code 构建的长篇网文创作辅助系统，旨在通过 RAG（检索增强生成）技术解决 AI 写作中常见的上下文遗忘与内容幻觉问题。该项目通过一系列自定义指令（Skills）与 Agent 协同工作，支持从大纲规划、内容撰写到章节评审的全链路流程，并能处理高达 200 万字量级的连载需求。此外，系统还内置了一个可视化 Dashboard，能够直观展示实体图谱、项目状态及“追读力”指标，为长篇小说创作提供系统化的数据支撑。
该项目特别适合网文作者及致力于长篇小说创作的 AI 爱好者，能够大幅降低复杂长篇叙事的逻辑维护成本。其热度源于对 AI 写作痛点的精准解决，特别是其引入的“追读力”系统和混合检索模式，将工业级的自动化创作能力带入了个人创作领域。对于追求高效创作且希望利用 AI 进行大规模文本连载的创作者而言，该工具提供了从环境部署到创作管理的完整解决方案，极大提升了创作的连贯性与专业度。
7. openai/skills Not new today
Skills Catalog for Codex
First seen: February 04, 2026 | Streak: 1d
Analysis
The `openai/skills` repository serves as a centralized catalog for "Agent Skills," which are standardized sets of instructions, scripts, and resources designed to enhance the capabilities of AI agents within the Codex ecosystem. By packaging specific task-oriented functionalities, these skills allow developers to create repeatable workflows that can be seamlessly integrated across different environments. Technically, the system utilizes a command-line interface, `$skill-installer`, to fetch and implement curated or experimental skills directly from the repository, requiring only a simple restart of Codex to activate the new logic.
This project primarily benefits software engineers and AI developers who seek to streamline complex automation tasks and improve the modularity of their AI-driven applications. It is trending because it addresses the growing demand for an open standard in agentic workflows, enabling teams to share and adopt standardized toolsets for diverse technical requirements. By fostering a collaborative ecosystem where custom functionalities can be easily distributed and reused, this repository significantly lowers the barrier to deploying specialized AI agents in professional development environments.
8. TheCraigHewitt/seomachine Not new today
A specialized Claude Code workspace for creating long-form, SEO-optimized blog content for any business. This system helps you research, write, analyze, and optimize content that ranks well and serves your target audience.
First seen: March 05, 2026 | Streak: 3d
Analysis
SEO Machine is a specialized Claude Code workspace designed to automate the end-to-end lifecycle of long-form, SEO-optimized business content. It leverages a suite of custom commands and autonomous agents to handle complex tasks, including competitive research, keyword mapping, content drafting, and data-driven performance analysis. Technically, the system integrates with external tools like Google Analytics, Google Search Console, and DataForSEO to provide real-time insights, while requiring users to define a "context-driven" framework via structured directories that enforce brand voice, style, and SEO guidelines.
This project is an ideal solution for content marketers, SEO strategists, and SaaS founders who want to scale their content production without sacrificing quality or ranking potential. It is gaining traction because it bridges the gap between generic AI generation and professional editorial standards, effectively transforming Claude Code into a comprehensive, agentic marketing team. By automating the technical overhead of SEO audits and internal linking, the system allows users to focus on strategy and growth rather than the manual labor of content optimization.
An AI Hedge Fund Team
First seen: March 07, 2026 | Streak: 1d
Analysis
The AI Hedge Fund project is an educational proof-of-concept platform that simulates an AI-powered investment firm by utilizing a multi-agent system. It integrates specialized "investor agents," modeled after legendary figures like Warren Buffett and Michael Burry, alongside dedicated functional agents that perform sentiment analysis, valuation, and technical analysis. Technically, the system functions by aggregating financial data and using LLMs to synthesize these diverse perspectives into cohesive trading signals, all managed by a centralized Portfolio Manager and Risk Manager.
This project is primarily aimed at developers, finance enthusiasts, and students who want to explore the intersection of generative AI and quantitative finance. It is currently trending because it provides a highly accessible, modular framework for experimenting with agentic workflows in a domain as complex as market analysis. By offering both a command-line interface and a web application, the project democratizes complex investment modeling while serving as a valuable learning tool for those interested in autonomous AI decision-making.
Select context for coding agents directly from your website
First seen: March 07, 2026 | Streak: 1d
Analysis
React Grab is a development utility that streamlines the workflow between a running web application and AI-powered coding agents by capturing precise component context. By simply hovering over a UI element and pressing a keyboard shortcut, the tool copies the associated file path, React component structure, and HTML source code directly to the clipboard. Technically, it integrates into development environments via a simple script injection—supporting frameworks like Next.js, Vite, and Webpack—and exposes a robust API for building custom plugins or leveraging low-level primitives for advanced inspection tasks.
Developers and teams utilizing AI coding assistants like Cursor, Claude Code, or GitHub Copilot will benefit most from this project, as it significantly reduces the friction of providing accurate project context. By automating the extraction of component metadata, it allows these agents to generate more relevant and accurate code suggestions, reportedly increasing operational speed and accuracy by up to three times. The tool is trending because it bridges the gap between the visual UI and the underlying codebase, transforming the developer experience for anyone reliant on AI to navigate complex, component-heavy modern web projects.