AnyFlow - flow-matching video generation: training, inference and VBench evaluation for t2v and i2v models.

OpenAI-compatible proxy for DeepSeek R1 & V3 — plug DeepSeek into Cursor, Cline, Continue, or any OpenAI SDK. Local-first, streaming, tool calls.

Train JEV-style System 1 decision models on Qwen3.5 and Qwen3.8-27B: 24-slot decision head, KL distillation, calibration and serving. 在 Qwen 上训练 JEV 风格的 System 1 决策模型。

A virtual agentic office in 3D where four AI robot agents (Evan, Lummi, Athena, Zed) chat, collaborate and get work done. Zed, the reception manager, takes your commands and delegates tasks. Includes a live activity panel for agent status and permissions, a task calendar Zed runs automatically, and plugins like Gmail, Drive, GitHub.

CTF 逆向 AI 解题工作台（Electron / Windows）：AI 工具循环 + 本地验证闭环，MCP 接入 IDA/JEB/Burp；自动求解 APK/PE/ELF/PYC/DEX/JAR/Mach-O/Lua/Web-JS；内置 11 个方法论技能 + 54 页结构化知识库 + 经验自进化。

Measures which GGUF quant of a model to run on your GPU: real speed and quality loss (KL divergence), not estimates.

说出你的需要，让应用在身边的屏幕上诞生——云端编码 Agent 当场生成 Web 应用，投到电视或投影上，并用语音继续操作。

Developer and product skills for Claude Code, Cursor and Codex: 1-file AI evals and Chrome extension store assets.

Stanford CS336《从头构建大语言模型》2026 中配版：18 篇中文技术博客 + 19 篇「图-字幕」原始文稿，基于 video2blog-skill 整理。

🧠 Persistent local-first memory for AI coding agents — MCP server + CLI for Claude Code & Cursor. SQLite FTS5, secret redaction, project isolation.

连接 AI 与 Aseprite 的 MCP 服务，支持实时读取和编辑未保存画布、图像预览、撤销重做及基础动画，保留 CLI 文件批处理能力，让人与 AI 在同一文档中协作创作像素画；An MCP server connecting AI to Aseprite for collaborative pixel art, with live canvas editing, unsaved document access, image previews, undo/redo, basic animation, and CLI batch processing

Open-source Ask AI for your docs, with an MCP server for agents: cited answers on your own model key, from one static index built with your site. No vendor, no vector database. Starlight, Docusaurus and VitePress plugins, Next.js, or one script tag.

The harness that builds harnesses: tell it about your work and it builds a custom AI agent with its own tools, views, skills and UI. Runs in the browser, terminal, headless, or as a standalone app. Built on pi.

Control and observe a running Minecraft: Java Edition game from AI agents over MCP (Fabric and NeoForge, 1.21.1+)

The last agentic harness you'll ever need. Bring any model; Pennant does the rest on your Mac and iPhone, and asks before anything is sent, deleted or paid for. Free, Apache-2.0.

An assistant for Windows that works inside 118 desktop programs and services through code instead of screenshots. It writes each program's own files or calls its official interface, checks the result, and saves every change as a version. Works with 25 model providers or a local model.

Run stock llama.cpp with the stock Unsloth Qwen3.6-35B-A3B 4-bit on a 16 GB Mac.

Self-hosted shared memory for AI agents: vaults of people, facts and decisions your agents recall over MCP or REST, with semantic, keyword and graph search.

AI agents and workflows for Kotlin Multiplatform, as small typed graphs: loops, parallel steps, streaming, save points and human approval. Inspired by LangGraph.

Research and shared data contracts for Pick Your AI: component catalogs, compatibility rules, pricing, and evidence for AI stacks.

MVO WHATS PHP — biblioteca PHP nativa para WhatsApp, com QR Code, sessões persistentes, mensagens, mídia e integração MCP para LLMs.

Smriti (स्मृति): private, local-first agentic RAG with System 1 + System 2 thinking over your notes, books and documents. Guardrails, tracing and evals included.

Local, provider-agnostic agentic hub: an orchestrator and a squad of little monsters on your machine

Awesome AI Coding Tools (2026): Curated 3-tier taxonomy of coding assistants, CLI agents, local LLMs, MCP servers, and benchmarks

BharatCode — open source AI coding agent for VS Code, powered by bharatcode.ai. Fork of Kilo Code (Kilo-Org/kilocode), reconfigured for the BharatCode provider.

A local privacy firewall for AI agents: masks secrets and personal data before a request leaves your machine and restores them in the answer. Claude Code, Codex, OpenCode and OpenAI-compatible agents.

Memory for coding agents that checks itself against your code. Every memory carries evidence; stale facts never reach the context.

Run Qwen3.8-Flash-Next on ONE RTX 3090 (24 GB) + 64 GB RAM: 128K context, up to 2,100 tok/s prefill, 43–51 tok/s decode. vLLM runtime with hot MoE experts on the GPU and cold experts computed on the CPU, INT8 KV cache, Docker, OpenAI-compatible API.

Run coding agents in a Linux VM on a clone of your project, then review and apply their changes like a pull request

Tidytalk：Mac 本地语音输入工具。按住说话，松开即可得到整理后的文本。语音识别完全本机运行，可搭配 Ollama 本地大模型润色转写内容，数据不会离开你的电脑。免费开源。| Private on-device voice input for Mac. Hold to speak, release to get tidied text. Local speech recognition, optional Ollama LLM polishing. No data leaves your computer. Free & open source.