Trending AI projects

AgentEnv Framework is a Python SDK/CLI for building, deploying, and scoring agentic environments (Docker or sandbox containers) that communicate via a standard protocol, enabling easy evaluation of LLM‑style agents in defined tasks.

A collection of ComfyUI workflows for MiniMax H3 Director, enabling text-to-video, image-to-video, and advanced video-to-video editing with reference materials.

Overalls is a self-hosted developer platform that provides persistent, AI-ready virtual machine environments for coding and automated agent workflows.

A security-first MCP server that gives LLM agents controlled SSH access to remote hosts, with command classification, policy-based authorization, human approval for destructive actions, and full audit logging.

rayskills is a toolkit of 8 specialized AI skills for agents like Claude Code that automates professional workflows such as content production, knowledge management, and publishing while enforcing safety boundaries.

LibreDesk is an open‑source, self‑hosted help‑desk that combines email, live‑chat and WhatsApp into a single inbox. It offers AI‑powered answer suggestions and a knowledge‑base‑driven assistant, plus typical support‑team features like automations, SLA tracking, SSO and API/webhook integrations. Deploy via Railway, Docker, or a downloaded binary.

An open-source framework that enables coding agents to generate production-ready Lottie animations from text prompts and SVG assets.

A minimalist ML framework for Rust that enables lightweight, Python-free serverless inference and browser-based model execution.

An open-source proofreading software that detects grammatical and stylistic errors across 20+ languages, surpassing the capabilities of basic spell checkers.

A unified command-line interface that provides a consistent syntax for interacting with 11 different social media and web platforms, specifically designed to be easily integrated into AI agents.

A motion-first creative web engineering skill for AI agents and developers that uses real-world physics and automated headless verification to build high-touch interactive websites.

BAML is a programming language designed for AI agents that provides a strong type system and built-in evaluation tools to ensure reliable, structured outputs.

The project is a curated list of open-source LLM-based trading agents, MCP servers, and skills, organized by category, with selection criteria and notable entries.

Cotal is an open, NATS‑based pub/sub protocol that lets AI agents (Claude Code, OpenCode, Codex, etc.) coordinate in any topology—trees, graphs, swarms—using multicast, unicast, and anycast messaging. It provides presence, durable delivery, and per‑agent ACLs, with a CLI for setup, spawning agents, and a web dashboard. Cotal fills the gap left by MCP and A2A by offering a shared space for many agents to work together.

mcp-go is a Go library for building servers that expose data, tools, and prompt templates to AI applications via the Model Context Protocol (MCP). It handles protocol details so developers can focus on their tools. The library is under active development, implements MCP spec version 2025-11-25 (with backward compatibility), and supports resources, tools, prompts, async task execution, session management, and more.

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.

The source release of NVIDIA's open GPU kernel modules for Linux, enabling the compilation and installation of kernel-level drivers for Turing and later GPUs.

Caire is a content-aware image resizing library that uses seam carving to shrink or enlarge images while preserving important visual content like faces. It offers a command-line interface and supports features like face detection, masks, and concurrent processing.

ComfyUI‑MiniMaxH3Mod adds nodes that let you encode images, video, and audio into reusable `.safetensors` “RefMod” bundles for the MiniMax H3 model, load them, inspect token costs, and apply them as conditioning in ComfyUI pipelines.

An open-source GUI for yt-dlp that provides video downloading from 1,800+ sites with integrated AI video summaries, natural language editing, and AI-powered subtitle generation.

A physical ESP32-based dashboard that monitors Claude Code usage and rate limits via BLE, featuring mascot animations and HID shortcut buttons.

InfiniTensor is an open‑source C++ runtime that imports ONNX graphs, applies generic graph rewrites, and executes them on CPU or on custom accelerator back‑ends (InfiniOps/InfiniRT). It offers Python bindings, a test suite, and a modular design that lets hardware vendors plug in their own kernels.

NeuralScreen is a Windows‑only tool that applies NVIDIA’s DLSS 5 neural upscaling (and optional DLSS 4.5 frame generation) to the whole desktop or a single window in real‑time, offering sharper visuals and higher FPS for RTX 20‑series and newer GPUs.

BirdNET‑Go is a self‑hosted, real‑time sound‑analysis service that runs multiple AI models (BirdNET, Google Perch, BattyBirdNET, BirdNET Geomodel) to identify birds, insects, mammals and bats from live audio streams. It offers a web UI with live spectrograms, a consensus‑based detection engine, extensive alert integrations, and can run on a Raspberry Pi or in Docker, all with local‑only inference.

A testing utility that uses the Codex CLI to benchmark model performance and reasoning token usage on a standardized math problem.

Voxenra is a cross-platform medical imaging workstation for DICOM files that integrates professional viewing, 3D reconstruction, and local AI-powered organ segmentation.

An official Rust SDK for the Model Context Protocol (MCP) that enables developers to build interoperable AI tools, resources, and prompts using Rust.

An automatic hyperparameter optimization framework for machine learning that uses a define-by-run API to dynamically optimize model settings.

An AI agent skill that provides access to a curated library of 10,000+ image generation prompts, allowing users to search for or generate custom prompts based on their specific content needs.

herdr‑annotate is a Rust plugin for the Herdr terminal AI‑assistant that captures selected terminal text, formats it as numbered Markdown annotations, and lets you copy, archive, paste, or send those notes directly to the active LLM. The Full variant also includes a TUI (plannotator‑tui) for reviewing Markdown files and the agent’s previous replies. Installable on macOS, Linux, and Windows.

A highly efficient gradient boosting framework that uses tree-based learning algorithms to handle large-scale data with high speed and low memory usage.

An agentic application security pipeline that automates the discovery, adversarial verification, and remediation of source code vulnerabilities to reduce the time from disclosure to validated fix.

A lightweight, generalist framework for Named Entity Recognition (NER) that enables zero-shot extraction of any entity type without requiring task-specific training data.

An open-source AI agent that automates Android and iOS devices using natural language to perform tasks and scrape data from apps.

A cloud-native platform for managed, team-owned AI agents with persistent work environments and shared knowledge across multiple communication channels.

myAiVideos is an open‑source, Docker‑compose‑based system that automates the whole short‑video creation workflow (topic → script → visuals → audio → post‑production → distribution) for Chinese content. It uses a FastAPI orchestrator plus a Java Spring‑Boot gateway for auth, routing, circuit‑breaking and metering, and integrates multiple LLM and multimodal models (DeepSeek, GLM, Kling, TTS services). The pipeline is layered, fail‑over capable, and fully observable via Langfuse and Prometheus.

Exo is an open‑source, self‑hostable runtime that lets you write LLM agents in markdown, choose any supported harness (e.g., Codex, Claude Code, Pi), and run them with persistent state, sandboxed execution, and secure credential handling.

GLM‑5 is an open‑weight LLM family (up to 744 B parameters, 1 M‑token context) focused on coding, cyber‑security, and long‑horizon agentic tasks. It introduces sparse‑plus‑linear attention (Flash), the IndexShare efficiency trick, and a reasoning‑effort API. Models are downloadable from Hugging Face/ModelScope and can be served with SGLang, vLLM, Transformers, KTransformers, or Ascend NPU. Fine‑tuning is supported via the slime RL framework and ms‑swift. The series claims state‑of‑the‑art open‑source performance on several benchmarks.

Lance is an open-source lakehouse format optimized for multimodal AI, providing high-performance vector search, fast random access, and efficient data management for ML workflows.

This repository hosts the official collection of ComfyUI workflow templates and reusable subgraph blueprints, providing a standardized way to share, browse, and deploy complex node-based workflows.

A Unity‑official plugin that equips AI agents (Claude Code, Codex, Grok, Antigravity) with natural‑language skills to edit Unity projects—adding UI, ads, assets, and more—by calling the Unity CLI. Install via each agent’s marketplace commands or manually, then invoke skills with `/unity:` or by describing the task.

Humming is a lightweight, high-performance JIT-compiled library for quantized matrix multiplication, enabling efficient inference across a wide range of low-bit precision formats on NVIDIA GPUs.

A collection of standardized task definitions and instructions that enable AI coding agents to perform complex AI/ML workflows on the Hugging Face Hub.

An in-process vector search library for Python and Rust that uses TurboQuant to compress embeddings by up to 15.5x, enabling fast, memory-efficient similarity search without the need for a dedicated database server.

A secure, validated registry and CLI for installing verified skills and instructions to extend the capabilities of professional AI coding agents.

Agentero is a cross‑platform desktop app that lets researchers read PDFs, take linked Markdown notes, and interact with local AI agents via open ACP/MCP protocols. It extracts equations and tables into structured Markdown/LaTeX, integrates Zotero libraries, supports bulk paper discovery (Cool Papers, RSS, arXiv), and offers a headless CLI for automation. All data stays in plain files and can be synced via S3 or SSH. MIT‑licensed.

Dynamo is an open‑source, Rust‑based orchestration layer that coordinates multiple GPUs or nodes for LLM inference. It adds disaggregated pre‑fill/​decode, KV‑aware routing, cache off‑loading, auto‑scaling and Kubernetes integration on top of engines like vLLM, SGLang or TensorRT‑LLM, delivering up to 7× higher throughput and faster cold‑starts for large‑scale LLM serving.

GEO Citation Lab是一个公开实证研究工作台，提供大规模跨平台引用数据、61万条中文引用记录、研究论文集和在线测试，用于研究AI搜索引擎如何选择信源、吸收内容及呈现实体。