Trending AI projects

A self-hosted media manager that integrates searching, downloading, tagging, and organizing for music, movies, and TV, replacing multiple separate tools with a unified pipeline.

A self-hosted academic writing assistant that uses a two-stage AI process to polish and enhance text, featuring a built-in administrative dashboard for managing user access and usage limits.

Data‑Juicer is an open‑source, Python‑based data‑processing framework that provides 200+ modular operators and a scalable Ray execution engine for cleaning, deduplicating, and preparing multimodal data for foundation‑model training, RAG, and embodied‑AI pipelines.

verifiers is a Python library for defining RL‑style environments that train and evaluate large language models. It ships ready‑made tasks, integrates with Prime Intellect’s Environments Hub, the prime‑rl training framework, and a hosted training platform, and includes docs, tests, and contribution guidelines.

Agent Sessions is a macOS app that locally indexes and searches conversation histories from many coding‑assistant agents, lets you resume supported CLI sessions, and shows per‑session quota/price usage, all without sending data off‑device.

Roborev is a local, automated code review tool that integrates with AI coding agents to provide real-time feedback and auto-fixes for every commit.

Agents-A1 is a 35B Mixture-of-Experts agentic model that achieves trillion-parameter-level performance by scaling agent trajectories and distilling specialized domain expertise.

A Python package for statistical computation that provides tools for estimation and inference for a wide range of statistical models.

A cross-platform screen search tool that integrates screenshotting, offline OCR, translation, and reverse image search to extract and act on screen information.

LeetCUDA is an open-source book and repository of 200+ high-performance CUDA kernels designed to teach developers how to implement optimized GPU operations like FlashAttention and GEMM.

An open-source workbench for Generative Engine Optimization (GEO) that helps brands diagnose and improve their visibility in AI-powered search engines.

Teable is an open‑source, self‑hostable platform that blends a spreadsheet‑style UI with a PostgreSQL backend and built‑in AI agents. It offers AI chat, an “App Builder” that creates and deploys apps from natural‑language prompts, AI‑driven automations, and bulk field‑filling, all running in isolated containers. The product can be used as a hosted service or deployed on‑premise (full‑feature or data‑only modes). The community edition is AGPL‑3.0 (MIT for some packages).

A curated and regularly updated directory of third-party Claude API relay services, with monthly traffic stats, pricing comparisons, and verification tools to help developers pick a stable and cost-effective provider.

Thunderbolt is an open-source, cross-platform AI client that lets you choose your own models (frontier, local, or on-prem) and self-host the entire application to retain data control and avoid vendor lock-in. It is designed for enterprises that need on-prem deployment, with features like projects, skills, widgets, MCP support, and local-first storage with optional encryption.

Inspect Robots is an open‑source Python framework for defining robotics benchmarks and running any policy—LLM agents, VLAs, scripted scripts—on real robots, simulators, or a mock world. It logs full sensor streams, policy transcripts, and scores, and offers live Rerun visualisation. Plugins add support for specific robots (YAM, Franka, etc.) and backends (ROS, Isaac Sim). The core is lightweight (NumPy only) with optional extras for visualization, LLM agents, code‑as‑policy, and voice input.

A privacy-first, open-source Android app that consolidates multiple AI assistants into a single, customizable interface with built-in ad and tracker blocking.

A real-time screen translator that uses OCR and machine learning to translate text from PC games and applications directly on the screen.

MoAI‑ADK is a Go‑based harness that orchestrates Claude Code (and other LLM back‑ends) as verified coding agents. It introduces a leader‑lane factory model, a SPEC lifecycle (plan‑run‑sync), and strict evidence binding so agents can’t claim success without real test runs. Parallel lanes run in isolated Git worktrees, and a web console visualises progress. The tool targets developers who want trustworthy, cost‑controlled AI‑generated code.

HVE Core is Microsoft’s open‑source, AI‑augmented SDLC framework. It ships VS Code and Copilot‑CLI plugins that expose specialised agents, prompt‑based instructions and reusable “skills” to automate research, planning, coding, security checks and review. The system encodes company‑wide standards and compliance rules, letting AI carry context while humans retain judgment. Teams install the extension, invoke the RPI (Research‑Plan‑Implement‑Review) agent, and can fork/customise the repo for their own processes.

A collection of reusable AI assistant instructions and MCP servers that help AI coding tools understand and operate Microsoft Fabric workloads, APIs, and best practices.

Reef is an open-source infrastructure for building self-improving AI agents that learn from interaction feedback to update their own model weights or operational harnesses.

Tracefinity is a web app that uses AI (local ONNX models or Google Gemini) to automatically trace tool silhouettes from photos, lets you edit and store those shapes, and generates Gridfinity‑compatible 3‑D‑printable bins (STL/3MF/SVG). It runs in a single Docker container, supports ARM devices, optional GPU acceleration, and offers authentication, automatic naming, and multiple deployment options.

Ontop is a Virtual Knowledge Graph system that exposes relational database content as virtual knowledge graphs — data stays in the source databases. It translates SPARQL queries into SQL via R2RML mappings, optionally using lightweight ontologies. Built with Maven 3.6+ / Java 11; Apache 2.0 license; backed by the Free University of Bozen-Bolzano, Ontopic s.r.l., Birkbeck, and Southeast University.

A binary classifier that separates real e-commerce orders from fake 'brushing' orders (fake purchases meant to boost a seller's stats). The README gives no technical details beyond that.

A workflow engine for Claude Code that turns it into a multi-model orchestrator, allowing it to dispatch specialized work to models like Codex and Gemini while maintaining context.

Codex App Manager is a cross-platform desktop tool (Tauri v2 + Rust) that installs, updates, and uninstalls the official OpenAI Codex desktop app. It updates itself via a China-friendly mirror system (R2 globally + IHEP S3 in mainland China, with GitHub fallback), does incremental delta updates on macOS with signature verification and rollback, manages MSIX/portable builds on Windows, and is the official runtime for the awesome-codex-skins theming library. It never modifies Codex's own files. macOS builds are Developer ID signed and notarized; Windows installers are not yet Authenticode signed.

A macOS application for real-time audio transcription using Whisper and Parakeet engines, featuring flexible keyboard and mouse triggers for quick voice-to-text conversion.

An agent skill that extracts visual design identities from references into a machine-readable JSON "Design DNA" to ensure faithful and deterministic UI implementation.

A GPU-accelerated, lightweight learning library for robotics research that enables rapid prototyping of learning algorithms with minimal overhead.

tinygrad is a compact, open‑source deep‑learning library offering tensors, autograd, an IR‑based compiler, JIT execution, and support for many accelerators (CPU, CUDA, Metal, OpenCL, WebGPU, etc.). It sits between PyTorch’s full‑featured ecosystem and micrograd’s minimalism, exposing the compiler pipeline for easy hacking while still providing nn layers, optimizers, and data loaders for real model training.

Editable‑Design is a GitHub project that offers Codex Skills for AI‑driven creation of editable visual assets (posters, infographics, research diagrams). Users install a Skill, give a natural‑language prompt, and receive structured HTML/CSS (or PPTX) with real text, separate image layers, a visual editor, and a replay of the agent’s design steps.

A self-hosted enterprise knowledge management layer and MCP server that compiles organizational documents into a structured wiki for secure, permission-scoped AI access.

A comprehensive subtitle generation and editing workflow that uses AI for transcription and correction, offering a visual editor for fine-tuning and professional export options.

An AI-powered penetration testing framework that automates security assessments using autonomous agents, multi-agent orchestration, and MCP-integrated security tools.

OSWorld-V2 is a benchmark for evaluating AI agents' ability to operate real-world desktop environments using virtual machines and gated task datasets.

A declarative YAML-based framework that wraps SQL to create auditable, version-controlled dashboards and reports, specifically optimized for both human auditors and AI agents.

TypeWhisper is a macOS app that provides on‑device or cloud speech‑to‑text, global hot‑key dictation, and customizable AI workflows (translation, rewriting, extraction, etc.). It supports many local and cloud ASR engines, a plugin SDK, a local REST API, CLI tools, iCloud sync, and premium meeting‑automation features—all wrapped in a polished UI with visual shortcut editing and widget support.

A pair of Claude Code skills that enforce a spec-driven workflow: /spec designs a feature spec through clarifying questions, and /spec-impl implements it step by step only after you manually approve the spec. This prevents silent LLM improvisation, makes decisions explicit, and keeps every change reviewable.

A bridge that connects AI agents to Autodesk 3ds Max via the Model Context Protocol, enabling automated 3D modeling, material workflows, and scene management.

A high-performance, privacy-focused Windows screenshot and clipboard management tool that features local OCR, translation, and screen recording.

Mixar is an open‑source Blender 5.2 fork that adds an in‑app AI chat agent, layer‑based texture painting, and text‑/image‑to‑3D generation. The repo provides the full desktop client (Python addon + C++ overlay) and build scripts for Linux, macOS, and Windows. AI features require a Mixar account and communicate with a closed‑source backend; non‑AI Blender functionality works offline. Licensed under GPL‑3.0‑or‑later (mix of GPL‑2 for Blender parts).

SkillEvaluator is an open-source framework for validating and testing AI agent skills through a three-tier pipeline that covers security, semantic deduplication, and live performance evaluation.

A React component library built on shadcn/ui that provides pre-built UI elements like waveforms and orbs for building audio and AI agent applications.

An AI-powered IDA Pro plugin that assists binary reverse engineering through LLM-driven function explanations, automated renaming, and semantic knowledge graphs.

MTranServer is a cross‑platform, CPU‑only offline translation server that runs locally (desktop app, npm package, or Docker). It uses ONNX‑runtime models to provide fast English‑Chinese, German‑English, and other language pairs via a HTTP API compatible with DeepL, Google Translate, and various editor/browser plugins. After an initial download of the required model files, it can operate completely offline, offering a privacy‑friendly alternative to cloud translation services.

TorchLeet is a free collection of 68 real PyTorch interview problems (from Google, Meta, Anthropic, etc.) as Jupyter notebooks with question/solution pairs, organized into Basics, LLM Learning Path, and Advanced tracks, plus an MCP-based AI tutor that coaches without spoiling answers.

A video direction skill for Dreamina/Jimeng Seedance 2.5 that transforms creative ideas and reference images into structured video scripts and optimized prompts.

A user-friendly AI agent desktop application that provides personalized assistants with memory, personality, and the ability to operate a computer and browse the web.