Trending AI projects

A comprehensive architectural analysis of the Claude Code agent that reveals how to build reliable AI systems by focusing on deterministic infrastructure, safety, and context management rather than just AI logic.

OpenCluely is an open-source desktop assistant that provides real-time AI help during technical interviews via a stealth overlay that is invisible to screen-sharing software.

《Hermes Agent 白皮书》是开源社区第一本系统性的 Hermes Agent 中文指南，共 16 册、30 万字以上，覆盖安装部署、核心架构、记忆系统、技能工具、多平台接入、变现路径等完整知识图谱，帮助新手和开发者从入门到精通。

A lightweight proxy that allows using the Claude Code CLI and VSCode extension with various alternative, free, or local LLM providers instead of the official Anthropic API.

An open-source architecture diagramming tool that allows users and AI agents to create, edit, and synchronize diagrams with codebases via a shared DSL.

A high-performance GEMM kernel library for Huawei Ascend NPUs that abstracts complex hardware primitives to deliver near-peak computational efficiency for LLM training and inference.

A library of reusable, composable AI‑agent “skills” (prompt‑driven actions) for building autonomous software‑factory loops, installable via npm and usable with many coding assistants.

JevChat‑Windows is a Windows‑only desktop assistant that captures a chat window, runs offline OCR to read incoming messages, sends the text to a small decision model (Jev) and a generative LLM, and shows three ranked reply suggestions in a floating overlay. Users paste the chosen reply themselves. It works with WeChat and KakaoTalk, stores API keys only in the registry, and keeps all screenshots in memory. The project is MIT‑licensed (with a GPL‑v3 UI component) and provides a ready‑to‑run executable as well as source‑code for developers.

A Rust-based reverse proxy that transforms the free DeepSeek web interface into a standard, compatible API for OpenAI and Anthropic clients.

SHARP is a fast monocular view synthesis tool that uses a neural network to generate a metric 3D Gaussian representation from a single image in under a second.

Supergateway is a CLI tool that converts MCP (Model‑Context‑Protocol) servers that normally speak over stdio into web‑friendly transports (SSE, WebSocket, Streamable HTTP) and vice‑versa. It can also wrap remote MCP endpoints, combine multiple servers via a JSON‑C config, and add security, health checks, and logging. Ideal for exposing AI‑tool servers to browsers or remote debugging without rewriting the server code.

UE‑MCP is a Node.js server + C++ bridge that lets AI assistants control Unreal Engine 5 projects via thousands of editor actions, supporting natural‑language commands, YAML workflow scripts, and extensible npm plugins, with safety guards and cross‑platform support.

A vision toolkit for DeepSeek Harness that gives text-only agents the ability to perform image Q&A, UI restoration, and GUI automation through a combination of vision models and local processing tools.

FigMirror is a tool that replicates the visual style of scientific figures from research papers and applies it to new data, generating editable matplotlib code.

SkillAlchemy is an open‑source framework that automatically builds reusable, installable agent skills from vague task briefs by mining public documentation, papers, and code. It discovers hidden requirements, extracts evidence‑backed procedures, judges their scope, and packages them for LLM agents like Claude Code or Codex. Evaluated on 87 benchmark tasks, it reaches a 55.8 % pass rate, outperforming existing automated skill creators and rivaling human‑curated skills. Install via `npx skills add agentsope/SkillAlchemy` and invoke the agent to generate a skill.

A GEP-powered self-evolution engine for AI agents that transforms runtime logs and experiences into auditable, reusable evolution assets to replace ad hoc prompt tweaking.

A step-by-step educational project that implements a minimal coding agent from scratch in TypeScript and Python to explain the architecture of AI coding assistants.

NVIDIA NeMo DataDesigner is an open‑source Python library for building synthetic‑data pipelines. It lets you define columns (categorical samplers, LLM‑generated text, image/audio generators), specify cross‑field dependencies, validate rows with Python/SQL/LLM judges, and run the whole workflow via a CLI or Python API. Plugins add extra data sources and generation capabilities. The tool is model‑agnostic (supports NVIDIA Build, OpenAI, OpenRouter, etc.) and is aimed at creating production‑grade multimodal datasets for AI training.

claude‑real‑video (crv) is a Python CLI / web tool that extracts smart keyframes and Whisper transcripts from any video (YouTube, local file, etc.) on your own machine. It deduplicates near‑identical shots, optionally adds speaker labels, and outputs frames, timestamps, and a transcript that you can paste into Claude, ChatGPT, Gemini, or other LLMs. The package also provides a local knowledge‑base (`crv‑ask`) and integrates with Claude Code, Cursor, Codex, and other agent platforms via a one‑line skill install.

An agentic second brain that uses AI agents and local Markdown files to automate knowledge synthesis, professional workflows, and personal organization.

Surf is a command-line tool that lets AI agents control Chrome via simple commands and a browser extension, offering zero config, agent-agnostic use, and the ability to query AI services without API keys.

An information processor for quantitative finance that transforms raw papers and news into structured, cited, and timestamped financial knowledge for RAG and agentic reasoning.

An open tabular foundation model that uses in-context learning to provide high-performance classification and regression without requiring hyperparameter tuning.

Open Mercato is a TypeScript/Next.js/PostgreSQL framework that ships with ready-made business modules and a structured environment for AI coding agents, so you start with ~80% of a working back office and only build the rest (the 20% specific to your business) with the help of AI. It includes agent-facing rules, specs, and an optional orchestrator (Cezar) for unattended tasks.

KTransformers is a research framework for efficient LLM inference and fine-tuning via CPU-GPU heterogeneous computing, enabling large MoE models like DeepSeek-R1 to run on a single consumer GPU (24GB VRAM) with 3-28x speedups, and offering fine-tuning with 6-12x speedup over ZeRO-Offload.

An OBS Studio plugin that uses neural networks to remove portrait backgrounds and enhance low-light video in real-time without requiring a physical green screen.

An industrial-strength NLP library for Python and Cython that provides pretrained pipelines and neural network models for advanced text processing across 70+ languages.

A Python library and runnable textbook providing 35 production-grade agentic AI patterns with a unified API and multi-provider support.

LLM Gateway is an open‑source API layer that lets you call many LLM providers through a single OpenAI‑compatible endpoint, with built‑in key management, cost tracking, and performance monitoring. It can be used as a hosted SaaS or self‑hosted via Docker.

dsh‑commandcode‑provider is an unofficial DeepSeek Harness plugin that adds Command Code as an LLM provider. It offers UI integration, multi‑account rotation, plan‑aware model picking, vision support, cost estimates, quota display, web‑search, and optional zero‑data‑retention, working with dsh 0.2.x.

An AI-powered database client and SQL workspace that allows developers and analysts to generate, explain, and optimize SQL using natural language.

A Python library providing a unified interface for time series analysis, including forecasting, classification, and clustering, compatible with scikit-learn.

An open-source AI camera platform that enables local deployment of VLM scene analysis and object detection skills with autonomous, hardware-aware installation.

A machine learning library that parses raw PDF scientific publications into structured XML/TEI documents for high-scale bibliographic data extraction.

An autonomous offensive security framework that automates bug bounty hunting and vulnerability research using an AI agent or MCP server integration.

A 3D visualization tool that replays coding-agent sessions on a map of a codebase to reveal how agents explore and modify files.

An open-source PyTorch toolkit for Conversational AI that provides recipes and tools for speech and text processing, including speech recognition and synthesis.

A community-curated library of 6,582 text prompts for ByteDance's Seedance 2.0 video generator, with 6 featured examples, sample videos, author credits, a companion web gallery, and a CC BY 4.0 license.

An open-source, general-purpose AI agent with durable memory and a self-evolving core that can rewrite its own code and architecture to improve itself.

numbat is an endpoint visibility and security tool that monitors, detects, and optionally blocks unauthorized AI agent activity across desktops, CLIs, and IDEs.

A self-hosted mission control dashboard that manages multiple AI coding agents in persistent tmux sessions, providing a mobile-optimized web interface for remote monitoring and control.

llm-wiki-compiler (llmwiki) is an open‑source CLI/SDK that compiles raw documents into a citation‑aware markdown wiki. It enforces a declarative profile (CLP), supports review gates, offers hybrid retrieval, exports via Open Knowledge Format, and provides an MCP server and TypeScript SDK for agent integration.

libvips is a fast, memory-efficient image processing library with ~300 operations, support for many formats (JPEG, PNG, TIFF, WebP, HEIC, PDF, SVG, etc.), bindings for 11+ languages, and adoption by projects like Mastodon, sharp, and Ruby on Rails. It's LGPL-licensed, installable via package managers or built from source with Meson, and has many optional dependencies for format-specific support.

Claude of Tanks is a browser‑based, Three.js‑powered tank combat simulator featuring 220 procedurally generated vehicles, 33 destructible battlefields, 13 game modes, and WebRTC multiplayer, all built with a Claude/Codex AI‑assisted pipeline.

BambooAI is an LLM-driven data analysis platform that prioritizes transparency and reproducibility by maintaining an inspectable, executable record of every computational step taken during an investigation.

A cost-effective computer-use agent that replaces expensive frontier-model planning with a fast, low-cost classifier for desktop automation.

Tsec-Hackathon is the official resource hub for Tencent Cloud's Intelligent Penetration Hackathon, aggregating defense slides, event materials, and open-source repositories from the top teams across two editions. It serves as a one-stop navigation page for learning how LLM-based penetration testing agents are built and evolved.

PowerInfer is a CPU/GPU hybrid LLM inference engine that uses activation locality to run large models quickly on consumer-grade GPUs by offloading 'cold' neurons to the CPU.