A custom vLLM-style inference engine implementation that benchmarks Flash Attention and Paged Attention to optimize LLM prefilling and decoding.

An autonomous agent framework that runs unattended on GitHub Actions to ship code, deploy apps, and find vulnerabilities using a prompt-based skill system.

A framework for self-improving LLMs that evolves system prompts into structured playbooks through a modular process of generation, reflection, and curation.

A unified storage SDK that provides a single API for interacting with multiple object and blob backends like S3, GCS, and Azure.

tiny‑vllm is an open‑source C++/CUDA LLM inference engine (with optional AMD‑HIP support) that loads a Llama 3.2 1B model from Safetensors and implements the full forward pass—including custom kernels for embeddings, RMSNorm, RoPE, grouped‑query attention, Flash‑style softmax, and paged KV‑cache. It doubles as a step‑by‑step tutorial covering every component of transformer inference, aimed at learners and educators.

A Godot Engine addon that enables developers to create complex, adaptive game AI using behavior trees.

A universal infrastructure for autonomously evolving AI agents by mutating their workspace files (prompts, skills, and memory) to improve performance across any domain.

A versatile 3D mapping framework that uses factor graphs and GPU acceleration to provide accurate, sensor-agnostic range-based mapping and localization.

An agent-workflow-optimized GitLab MCP server that enables AI agents to manage projects, merge requests, and pipelines via over 260 granular tools.

An on-device AI agent for Android that integrates chat, system-level tools, and a local workspace to execute complex tasks directly on the device.

Deep Eye is a Python CLI that orchestrates multiple LLM providers to automatically generate, fire, and triage over 50 web‑application vulnerability checks, producing compliance‑mapped bug‑bounty reports.

A secure Telegram client for the OpenCode CLI that allows users to remotely run AI coding tasks, manage sessions, and monitor progress from their phones.

A virtualization project that enables running full macOS on M1/M2 iPads, allowing the use of pro apps like Xcode and Terminal.

A development playground and toolkit for the Neuro SAN framework that enables the rapid design and deployment of declarative, multi-agent AI systems.

A cloud-native, one-stop MLOps platform that manages the entire AI lifecycle from data labeling and distributed training to model deployment, with extensive support for heterogeneous domestic AI hardware.

An AI training analysis and metric observation platform that provides visualization, automatic logging, and experiment comparison for model training teams.

A remote access gateway and visualization tool that lets you manage local AI agent CLIs and workstation data through a web browser from any device.

A collection of demonstration applications showcasing various ways to build AI-powered agents and apps using the Claude Agent SDK.

An LLM-routing coding agent that dynamically assigns development tasks to the best-suited models from providers like Gemini, ChatGPT, and Claude.

A local deployment skill for MiniMax H3 that enables low-VRAM NVIDIA GPUs on Windows to generate videos with native audio via AI agents.

A strong, open-source UCI Xiangqi engine derived from Stockfish that uses NNUE neural networks to analyze positions and compute optimal moves.

A local-first, open-source WYSIWYG Markdown editor with native AI integration for polishing, rewriting, and expanding content while keeping data on the user's disk.

A token-efficient MCP server that allows AI agents to interact with the Cloudflare API using server-side code execution to avoid context window exhaustion.

An AI-powered image translator that detects, removes, and replaces text in manga and comics, supporting multiple languages and automated typesetting.

A real-time object detection framework based on the Detection Transformer (DETR) architecture that aims to outperform YOLO models in speed and accuracy.

KataGo is a superhuman open-source Go bot trained via self-play that provides advanced game analysis, score estimation, and support for multiple hardware backends.

An AI-powered BI platform that combines a natural language chat interface with a headless semantic layer to provide reliable, governed Text2SQL data analytics.

A skill installation platform for Claude Code, OpenCode, and Codex that provides a curated collection of academic and scientific AI capabilities.

An AI-powered long-form fiction engine that uses schema-driven cards, knowledge graphs, and automated workflows to maintain consistency and structure in massive novels.

A desktop application that converts screenshots, images, and PDFs into editable LaTeX formulas and text using OCR.

ExcelMcp is a Windows‑only server and CLI that let LLM agents control a live Microsoft Excel instance via its COM API, exposing 326 operations (Power Query, PivotTables, VBA, Python, etc.) for real‑time spreadsheet automation.

A cloud-native AI agent framework that decomposes monolithic agents into serverless components for better scalability, security, and cost-efficiency on AWS.

An AI-powered voice input method for Android that supports multiple cloud and local ASR engines with LLM-based text post-processing.

A self-hostable dashboard template that acts as a mission control center for the Hermes AI agent, allowing users to dispatch tasks and manage memory via a shared Postgres database.

An implementation of the Model Context Protocol for the Unity Editor, allowing AI assistants to directly manipulate scenes, assets, and editor settings.

A collection of on-device demo apps for iOS, Android, and HarmonyOS that allow the MiniCPM-V family of multimodal and text models to run fully locally via llama.cpp.

A synthetic patient population simulator that generates realistic but non-real health records in various formats like FHIR and CSV for testing and development.

A high-performance constrained decoding library that enforces JSON schemas, regular expressions, and context-free grammars on LLM outputs to ensure structured data.

AgentKit is an open‑source SDK that provides a catalog of 70 ready‑made “kits” (full apps, pipelines, or templates) for building, testing and deploying AI agents. Kits are defined by a simple config file, can be edited in a visual flow studio, and are shipped as Next.js apps that run server‑less. The collection covers support triage, code review, RAG chat, hiring assistants, legal/medical bots, and many other business‑focused agents, making it a turnkey platform for creating reliable, enterprise‑grade AI agents.

A modular, retrieval‑augmented prompt framework for AI‑assisted Chinese web‑novel writing. It splits the workflow into a short router (`SKILL.md`) and many self‑contained sub‑skills (concept planning, opening, plot logic, dialogue, consistency review, etc.). Each sub‑skill ships tutorials, runtime rules, good/bad examples, and may use local novel corpora via a provided search script. The system routes a user’s brief idea to the appropriate module, pulls relevant example excerpts, composes a concise prompt, generates text, and finally runs a consistency review. Designed for Chinese novel authors, prompt engineers, and researchers interested in modular LLM writing pipelines.

A specialized plugin for Claude Code and other AI assistants that provides 26 specialist agents and 51 skills to help developers build bug-free Elixir and Phoenix applications.

Warren is an open‑source, self‑hosted platform that turns LLM‑based coding agents into managed, observable workloads. It provides isolated workspaces, lifecycle control, spend limits, live event streams, automatic Git commits (and PRs), and supports local, Docker, or Kubernetes runtimes. Ideal for individuals or small teams that want to run coding agents securely on their own infrastructure.

An all-in-one Android automation and control platform that integrates UI automation, remote desktop, packet capture, and AI agents into a single unified service.

A desktop companion plugin for DeepSeek Harness that provides a real-time, visual status indicator of an AI agent's working state.

A universal, serverless pasteboard that enables real-time, end-to-end encrypted clipboard sharing across Windows, macOS, and Linux devices over a local network.

A configuration sidecar for Grok Bot that allows users to run any OpenAI-compatible model with verified wire-level mappings to ensure native performance and reasoning behavior.

FlashVSR is a one-step streaming diffusion framework for real-time video super-resolution that achieves high-speed upscaling and scalability to ultra-high resolutions.

An open-source MCP server that enables natural language interaction with Neon Postgres databases for tasks like project management, branching, and SQL execution.