Runnable, documented projects worth studying and extending.
OthersQwen-CodePWAn open-source terminal AI coding agent by the Tongyi Qianwen team, similar to an open-source alternative to Claude Code. Supports code generation, editing, debugging, and other operations within the terminal, based on the Qwen model.
SafetySherlockPWA username search tool spanning 400+ social platforms; enter your username to track your accounts across major social networks
OthersPraisonAIPWYour 24/7 AI workforce – a low-code multi-agent AI platform supporting automatic planning, research, and coding, with deliveries available via Telegram/Discord/WhatsApp
OthersOumiPWA one-stop platform for fine-tuning, evaluation, and deployment of open-source LLMs, supporting mainstream open-source models such as Qwen3 and DeepSeek-R1
OthersSuperset EditorPWIn the AI Agent era, code editors can run multiple Claude Code, Codex, and other coding agents simultaneously, managing and scheduling them uniformly to improve development efficiency.
OthersLiteParsePWThe LlamaIndex team has launched a fast open-source document parser supporting multiple formats including PDF/HTML/Office and providing high-quality structured data for RAG and AI applications.
Otherslast30days-skillPWAI Agent skill plugin can research any topic's activity across platforms such as Reddit, X, YouTube, HN, Polymarket, for nearly 30 days, generating verifiable comprehensive summaries. This week, GitHub added 10k+ stars.
OthersAIChatPWAn all-in-one LLM command-line tool integrating Shell assistant, Chat-REPL, RAG, AI tools, and Agent functions, supporting mainstream models such as OpenAI, Claude, Gemini, and Ollama.
Otherslast30days-skillPWAI Agent skill plugin can automatically research content from multiple platforms such as Reddit, X, YouTube, HN, Polymarket, and generate well-founded comprehensive summary reports.
AI and AgentsGoose — Block is an open-source scalable AI AgentPWBlock's open-source AI Agent is not just code suggestion—it can install, execute, edit, and test, supporting any LLM
OthersMulticaPWAn open-source AI Agent team management platform that turns coding agents into true teammates—assign tasks, track progress, and build skills. Supports multi-agent collaborative orchestration.
OthersMLX-LMPWApple's LLM inference tool in the MLX framework lets you efficiently run various large language models locally on your Mac
OthersHeadroomPWThe context optimization layer for LLM applications automatically compresses and manages context windows, reducing token consumption while maintaining output quality. Supports multiple compression strategies.
Othersmicrosoft/agent-frameworkPWMicrosoft's open-source AI Agent architecture, orchestration, and deployment framework supports Python and .NET, providing multi-agent workflow support and suitable for enterprise-level agent application development.
OthersSupersetPWIn the AI Agent era, code editors can run multiple Claude Code, Codex, and other coding agents locally simultaneously, with unified management and scheduling
OthersCarbonylPWChromium browser running in the terminal! Renders web pages as terminal character images, supports full web standards, and is suitable for GUI-free server environments. 18k stars。
Efficiency toolsOpenScreen - An open-source screen recording toolPWA free and open-source screen recording tool, watermark-free, subscription-free, and suitable for commercial use. An open-source alternative to Screen Studio, it ranked first on GitHub's trending list this week, with 15,000+ stars
OthersFireRed-OpenStorylinePWAI video editing agent transforms manual editing into intent-driven directing mode through natural language interaction. Supports LLM planning, precise tool orchestration, and reusable Style Skills, enabling transparent human-machine collaborative video creation.
OthersLiteParsePWAn open-source fast document parser by the LlamaIndex team, supporting multiple formats including PDF/DOCX/HTML, optimized for RAG pipelines, fast parsing, and preserving document structure
OthersOpen-LLM-VTuberPWAn open-source AI VTuber framework supporting hands-free voice interaction, voice interruption, and Live2D facial capture, capable of running locally and connecting to any LLM. Cross-platform support, suitable for creating personalized AI virtual streamers.
OthersAgentFieldPWAn AI backend framework that builds and runs AI agents just like microservices. Supports routing, orchestration, memory management, asynchronous execution, and encrypted audit traces, with each function automatically becoming a REST endpoint.
OthersHugging Face SkillsPWHugging Face's official Agent skill library allows your AI Agent to directly call models and tools from the HF ecosystem, supporting frameworks like smolagents
OthersAndrej Karpathy SkillsPWBased on Andrej Karpathy's observations on LLM coding pitfalls, the CLAUDE.md rule file helps Claude Code avoid common coding errors and improve the quality of AI-assisted programming. A single file can significantly improve AI coding behavior.
OthersDFlashPWBased on Block Diffusion, the fast speculative decoding framework accelerates LLM inference with diffusion models, significantly improving generation speed. Latest research findings from Z-Lab.