Trending AI projects

A self-hosted proxy that converts the free Google Gemini web interface into an OpenAI-compatible API, allowing users to access Gemini models without official API keys.

DeepSeek Whale-chan is an open-source community project that provides a complete character design specification, visual asset library, and agent skills to generate high-consistency illustrations, posters, stickers, and multi-panel comedy comics of the whale-fin maid character, across five head-to-body proportion forms, using advanced image generation models.

AutoSubs is a cross‑platform desktop app that generates offline subtitles from audio/video, offering speaker labeling, many local transcription models, and direct integration with DaVinci Resolve and Adobe editing tools.

An open-source, AI-integrated terminal that combines a command-line interface with context-aware AI assistance, durable SSH sessions, and built-in graphical editors.

A reusable agent skill that enables LLMs to build and maintain a structured wiki of synthesized knowledge, moving beyond simple RAG by compiling sources into durable, cross-linked markdown pages.

An MCP server that connects Google Search Console to AI assistants, enabling natural language analysis of SEO data, indexing audits, and sitemap management.

MCP Toolbox is an open‑source server that lets AI agents and IDEs talk to enterprise databases via the Model Context Protocol. It offers ready‑made database tools and a framework for defining custom, secure SQL/semantic‑search tools, with SDKs for Python, JS/TS, Go, and Java.

An engineering framework for AI coding that persists project specifications, tasks, and memory in the repository to ensure AI agents follow team standards.

A handbook that organizes 28 common reviewer critiques into actionable rebuttal strategies, including a reusable skill to analyze comments and draft responses, aimed at AI conference authors.

An open-source desktop application for generating and editing videos and images using LTX models, supporting both local GPU execution and cloud API modes.

A lightweight deep learning framework optimized for high-performance inference and training on mobile, IoT, and embedded devices.

A Model Context Protocol (MCP) server that connects AI assistants to official South Korean legal databases to provide accurate, verified legal research and legislative analysis.

Tracely is a self‑hosted observability and CI/CD platform for AI agents. It ingests OTLP traces, evaluates runs with built‑in and LLM judges, clusters failures, and lets you promote a real production failure into a hermetic regression test that blocks pull‑requests. The stack (API, worker, UI, ClickHouse, Postgres, Redis, MinIO) can be deployed with Docker‑Compose or one‑click on Railway. The `tracelyai-sdk` auto‑captures LLM calls, and the GitHub Action integrates the gate into CI.

Frametell is a macOS (Apple‑Silicon) desktop and CLI app that uses LLM agents to turn a sentence, notes, PDF, or slides into a narrated 1080p explainer video. It drafts a storyboard, writes and voices the script (ElevenLabs/Volcengine), draws each page in selectable styles, checks quality, and renders the final film—all locally, with support for many LLM providers.

A multi-stage serving runtime for omni, speech, and TTS models that optimizes generation across heterogeneous compute stages with an OpenAI-compatible API.

An open-source incremental indexing framework that keeps AI agent context fresh by only reprocessing data changes (deltas) from enterprise sources into vector or graph stores.

A tool that cleans up AI-generated pixel art by snapping pixels to a consistent grid and quantizing colors to a strict palette for game development.

Janas is a high-performance C-based inference engine that allows large mixture-of-experts language models to run on consumer hardware by streaming weights from disk and caching active experts in RAM.

A collection of code and datasets released by Google Research, providing public access to various research implementations.

Triton is an open-source language and compiler for writing fast custom deep-learning primitives, aiming for CUDA-level performance with higher productivity and flexibility. It supports NVIDIA and AMD GPUs on Linux, can be installed via pip or from source, and includes extensive debugging and development tools.

Jev Cookbook是Datawhale开源的Jev中文教程，以可运行的Jupyter Notebook教学，帮助开发者掌握TypeSafe System One模型（Choice/Score/Noul三种原语）的开发范式，包含实战配方、语音智能家居、模型评测与本地微调等完整内容。

n8n‑MCP is a Model Context Protocol server that exposes the full n8n node catalog (≈3 000 nodes) as an API. AI assistants can search nodes, fetch detailed property schemas, get documentation, pull example configurations, and retrieve ready‑made workflow templates. The service also offers validation tools and, when given n8n credentials, can create, update, test, and manage workflows directly. It’s designed to let LLMs reliably build and deploy n8n automations.

A comprehensive collection of Chinese-language study notes, translated subtitles, and supplemental materials for Andrew Ng’s classic Stanford Machine Learning course.

A collection of 35 paste-ready prompts and configuration files (system prompts, AGENTS.md files, Ollama Modelfiles) that re-shape model behavior for 25+ AI models, documented for research and education on prompt injection and instruction hierarchy.

Flix is an open-source, ESP32-based quadcopter built from scratch for education and research. It uses off-the-shelf parts, has clean Arduino firmware under 2,000 lines, communicates via MAVLink over Wi-Fi or ESP-NOW, and can be controlled by gamepad, RC transmitter, or smartphone. It includes a Gazebo simulator that runs the real firmware, a Python scripting library, and a planned position-control feature. The project provides full assembly instructions, wiring schematics, 3D-printable frame files, and a components list. It's a DIY project with no warranties — the README warns it's not easy to build and the author accepts no liability.

The official Swift SDK for the Model Context Protocol (MCP). It provides client and server components to build applications that communicate with AI models, supporting tools, resources, prompts, and more.

Meetingly is an open-source AI assistant that provides real-time answers, meeting transcripts, and screen analysis during video calls by capturing local system audio.

DirectorDesk is a cross‑platform desktop tool for building simple 3‑D stages, choreographing character actions, and designing camera moves. It exports reference videos (including depth maps) that can be fed to AI video‑generation models, and it offers a natural‑language AI assistant and support for external agents to automate edits.

Voyager is a cross‑browser extension that adds folder organization, prompt vaults, timeline navigation, export options, and UI tweaks to Gemini, Claude, ChatGPT, DeepSeek and other web‑based LLM chat sites.

A team-architecture factory for Claude Code that automatically transforms domain descriptions into structured multi-agent teams and skills.

Agent Governance Toolkit is a Microsoft‑maintained open‑source library that sits between autonomous AI agents and the tools they call, enforcing deterministic, auditable policies (allow/deny/require‑approval) written in YAML/OPA. It provides language‑specific SDKs (Python, TypeScript, .NET, Rust, Go), a CLI (`agt`) for compliance checks, and adapters for major agent frameworks, enabling production‑grade governance, attribution, and tamper‑evident logging.

A browser bridge and extension that connects DeepSeek Harness agents to active Chrome or Firefox tabs, allowing AI to operate the web while preserving user sessions and cookies.

Agnes Video Generator is an open‑source Python tool that creates multi‑scene AI videos from text or images. It uses the free Agnes AI cloud API for inference, so no GPU is needed locally. Features include TTS narration, auto subtitles, digital‑anchor avatars, image‑to‑video, keyframe animation, and unlimited free usage. Install via a simple `cp .env.example .env` and run locally, in Docker, via npm, or let an AI coding agent handle the setup.

A Codex Skill that converts reference images or AI-generated diagrams into native, editable Microsoft Visio (.vsdx) files for scientific paper figures.

A framework that assembles a multi-modal AI agent featuring persistent text-based memory, voice interaction, visualizers, and webcam-based gesture control.

SimpleEnglish is an Agent‑Skills plug‑in that forces LLMs to rewrite their replies in the disciplined, jargon‑free style of ASD‑STE100 Simplified Technical English. It works with any Agent Skills‑compatible model (Claude, Codex, Gemini, etc.), offers a plain mode and a stricter STE‑vocabulary mode, and is installed via a single `npx skills add` command plus an optional agent‑specific plugin. Benchmarks show a 95 % drop in formatting defects and a 78 % reduction in STE rule violations, with human judges preferring the skill‑enabled output in most cases. The project is MIT‑licensed and contains only paraphrased rule definitions.

A reference implementation for autonomous vulnerability discovery and remediation that uses Claude to find, verify, and patch security bugs in source code.

Awesome REA Cases is an unofficial, actively maintained collection of 817 reverse-engineering projects, experiments, and ideas where AI tools like Claude and REA help revive old devices, rebuild apps, preserve games, and understand hidden behavior. Each case links to public evidence and notes limits, updated six times daily.

FinGPT is an open‑source suite of financial large language models fine‑tuned with LoRA/QLoRA for tasks like sentiment analysis, relation extraction, headline classification, NER, QA, and stock‑price forecasting. It offers multi‑task and single‑task checkpoints, public instruction‑tuning datasets, a benchmark suite, and demo interfaces, all under an Apache‑2.0 license.

goofish-cli is a Python CLI that automates Xianyu (Chinese second‑hand marketplace) operations—publishing items, managing listings, sending/receiving messages, searching, etc. It outputs data in JSON/YAML/table formats, includes rate‑limiting and risk‑guard checks, and registers as an MCP server so AI agents (Claude Code, Cursor, OpenClaw) can call its commands as tools. Five Claude skill bundles are shipped for zero‑shot‑free agent use. The package is pip‑installable, open‑source under Apache 2.0, and comes with unit tests and documentation.

Genjutsu is a creative coding plugin for AI agents that enables the generation of high-quality, interactive UI/UX and full websites while preventing generic AI design patterns.

An open-source, local-first AI research workbench that enables reproducible science through scientific AI agents, integrated Python/R execution, and traceable artifact provenance.

A server that enables AI agents to interact with LinkedIn by automating browser actions, allowing them to read profiles, search for jobs, and manage messages.

A comprehensive educational repository providing a structured curriculum and practical code examples for software engineers to master AI integration, autonomous agent development, and AI-driven operations.

A collection of reusable AI agent skills that enable coding assistants to perform specific Unity development tasks like project configuration, asset management, and editor automation.

Potpie creates a living context graph from your codebase and development tools, providing AI agents with the deep project knowledge required to plan, debug, and write code.

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 comprehensive roadmap for Forward Deployment Engineers, detailing the technical skills, AI agent frameworks, and consulting strategies needed to deploy complex software in real-world, high-stakes enterprise environments.