Trending AI projects

A visualizer for Claude Code and Codex sessions that transforms text transcripts into interactive, live flow graphs for monitoring and reviewing agentic workflows.

A search-first, profile-aware MCP server designed for front-end and binary reverse engineering, allowing AI agents to perform complex analysis without being overwhelmed by tool schemas.

A bridge that connects AI agents to Cheat Engine via the Model Context Protocol, enabling natural language interaction with program memory for reverse engineering and modding.

Miles is a PyTorch‑native, enterprise‑grade reinforcement‑learning framework for post‑training trillion‑parameter models. It combines SGLang for ultra‑fast rollout with Megatron‑LM (or FSDP2/torchtitan) for scalable training, offering async RL, low‑precision support, LoRA adapters, fault‑tolerant rollout routing, and day‑0 compatibility with frontier LLMs.

An AI-powered high-frequency trading bot for the Monad blockchain that uses the Jev model to capture the spread on the MON-USDC pair.

Token Tracker is a local‑first dashboard that automatically records token usage from 46 AI coding assistants, shows cost estimates, and offers a native desktop UI with widgets, a desktop pet, achievements, and optional cloud sync/leaderboard.

A tool that connects AI agents to your actual Chrome browser via an MCP server and extension, allowing them to perform tasks using your existing login sessions and site-specific knowledge.

Desktop tool for local music source separation via the pymss engine — manages models, runs separation jobs, provides a stem editor with mixing/export, batch workflows, and audio utilities (ASR, vocal-to-MIDI, SDR evaluation). Built with Tauri + Vue 3 + Python, licensed AGPL-3.0.

A C++20 graphics library for Metal 4 and Vulkan 1.4 that replaces traditional resource binding with GPU pointers and descriptor heaps for a more intuitive data model.

PaperForge is a prompt-and-skill package that turns AI assistants into deep, critical paper readers. It guides a 13-step analysis that reconstructs the authors' reasoning, explains methods mechanistically, stress-tests assumptions, and proposes novel follow-up research ideas.

A load balancer and proxy for ChatGPT accounts that enables request distribution, usage tracking, and API key management via an OpenAI-compatible interface.

DevSpace is a self-hosted MCP server that securely connects ChatGPT to your local machine, letting it read, edit, search, and run code in your own projects. You run it locally, expose it through a tunnel you control, and approve connections with a password. It supports Linux, macOS, and Windows (with Bash-compatible shells), requires Node >=22.19 <27, and is installed via npm.

A collection of Jupyter notebooks and exercises designed to teach machine learning fundamentals using Scikit-Learn and PyTorch as a companion to the O'Reilly book.

An open-source Python framework designed for security professionals to proactively identify risks and vulnerabilities in generative AI systems through red teaming.

A Krita plugin that integrates generative AI diffusion models into the painting workflow, offering tools for inpainting, live painting, and precise structural control.

An agent skill that uses a closed-loop cycle of image generation and AI critique to build visually impressive games, apps, or 3D scenes.

A macOS video editor with built-in generative AI and MCP support, enabling AI agents to create and edit content directly on the timeline.

GLM‑OCR is a compact (0.9 B) multimodal OCR model for complex document understanding. It achieves top‑rank performance on benchmarks, runs efficiently via vLLM/SGLang/Ollama, and offers a Python SDK, CLI, and Flask server. You can use it through Zhipu’s cloud MaaS API (no GPU) or self‑host the model with optional layout detection. The project is open‑source (Apache 2.0 code, MIT model) and includes fine‑tuning guides and deployment examples.

An enterprise-grade TypeScript framework for building production-ready MCP servers with built-in authentication, dependency injection, and visual debugging tools.

Rainman Translate Book is an agent‑skill that converts a PDF/DOCX/EPUB into Markdown, splits it into chunks, translates each chunk in parallel with LLM sub‑agents (Codex, Claude Code, OpenClaw), validates via a SHA‑256 manifest, merges the results and builds HTML/DOCX/EPUB/PDF outputs, offering resumable runs and a glossary for consistent terminology.

An open-source Python framework for building long-horizon agents with a durable runtime to ensure reliability and resumability in production environments.

A curated collection of 17,979 creative prompts for OpenAI's GPT Image 2, with a web gallery, categories, and featured examples.

Peekaboo is a macOS‑only CLI and menu‑bar app that captures the screen, builds a structured UI map via accessibility APIs, and lets you automate clicks, typing, key presses, scrolling, etc. It also includes an LLM‑driven “agent” that can execute natural‑language instructions, and it can be exposed to external Model‑Client‑Protocol (MCP) clients such as Codex or Claude Code.

A cross-platform CLI that enables AI agents to observe and act on iOS and Android screens via token-efficient UI outlines and simplified interaction commands.

A cross-platform machine learning accelerator that optimizes inference and training for models from various frameworks like PyTorch and TensorFlow.

Claude Octopus is a Claude Code plugin that lets you invoke up to twelve external LLM providers for structured multi‑model reviews, council deliberations, and autonomous spec‑to‑code pipelines. It stays dormant until you run `/octo:*` commands, offers 31 built‑in personas, persistent memory, safety gates, and works on Linux/macOS (Windows via WSL). MIT‑licensed and actively maintained.

SourceWeft is a self-hostable, open-source AI workstation that combines chat, knowledge sources, files, and task execution. It uses multi-agent collaboration, a skill library, MCP tool integrations, and sandboxed execution, while supporting bring-your-own models and API keys. It offers desktop clients for macOS, Windows, and Linux plus a web app, and supports team collaboration with roles and access controls. Licensed under Apache 2.0.

APM is a Microsoft‑backed, open‑source dependency manager for AI coding agents. It lets you declare skills, prompts, plugins, and MCP servers in a single `apm.yml` manifest, resolves transitive dependencies, locks versions with a hash‑based lockfile, and enforces security and policy via audits and optional `apm-policy.yml`. Installable via Homebrew, script, pip, WinGet, or Scoop, it targets agents like GitHub Copilot, Claude, Gemini, Cursor, and others.

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.

slime is an open‑source RL post‑training framework that couples Megatron‑LM training with SGLang inference, providing high‑performance, scalable, and customizable loops for fine‑tuning LLMs (GLM, Qwen, DeepSeek, Llama 3, etc.). It offers a unified data‑buffer architecture, Ray‑based orchestration, optional persistent storage, and extensive tooling for debugging, profiling, and fault tolerance. Used in the production of GLM‑4/5 series, it serves researchers and engineers building agentic or reward‑driven LLM systems.

groma.md is a tool that generates and maintains a live, interactive C4 architecture map of your repository stored as plain Markdown, ensuring your documentation stays in sync with your code.

CodeBoarding is an open‑source tool that statically analyses a codebase, clusters the code into high‑level components, uses LLMs to generate readable descriptions, and outputs an `analysis.json` that can be visualised as architecture maps in a web UI, VS Code, or CI. It helps reviewers see the impact of pull‑requests at the component level and can render documentation automatically.

A curated collection of adversarial jailbreak prompts and techniques targeting open-weight LLMs, documented with full prompt templates and verification screenshots for red-teaming research and defensive AI alignment.

MoviePy is a Python library that enables programmatic video editing, allowing users to cut, composite, and apply custom effects to video and audio files using code.

An ultra-lightweight, autonomous AI assistant infrastructure written in Zig that runs on minimal hardware with near-zero overhead.

A collection of starter settings, templates, and configuration scripts designed to improve Claude Code's memory and context retention through a structured memory bank system.

A conversational AI video creation tool that automates scriptwriting, media sourcing, and editing through natural language prompts.

A private, on-device voice-to-text and meeting transcription app for macOS that uses local AI models to ensure no data leaves the machine.

Vocero CRM is a self-hosted, open-source WhatsApp CRM that includes an AI agent and a testing laboratory to simulate and evaluate agent performance before it interacts with real customers.

A tool that creates persistent, isolated Linux workspaces using Docker or Podman to provide secure environments for software development and AI coding agents.

An MCP server that connects AI assistants to Grafana, enabling them to query observability data, manage dashboards, and handle alerts across multiple datasources.

TensorSharp is a .NET 10 inference engine for GGUF LLMs, offering CLI, web‑API and desktop/mobile (TensorAgent) interfaces, with CPU, CUDA, Metal and Vulkan back‑ends and support for many text, multimodal and generation models.

A high-performance document-intelligence engine that extracts clean text, tables, and structured data from over 100 file formats, including PDFs, images, and audio.

GodotPrompter is a plugin that equips AI coding agents (Claude, Copilot, Cursor, etc.) with a catalog of 56 Godot‑specific “skills”. When an agent receives a request (e.g., “add a player controller”), it loads the matching skill and returns Godot‑best‑practice GDScript / C# code, checklists, and explanations. It includes mentor mode, project‑scaffolding, gameplay systems, native extensions, and support for popular addons, all version‑locked to Godot 4.3+. Installation works via each platform’s plugin marketplace, and a SessionStart hook auto‑detects Godot projects.

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.

Banana Claude is a plugin that plans, reviews, and executes Google Gemini image‑generation requests, giving users a safe, cost‑controlled workflow for creating, editing, comparing and typesetting AI‑generated visuals.

The OpenAI CLI is an official command-line tool that allows users to interact with the OpenAI REST API directly from their terminal, enabling tasks like image generation, file management, and batch processing without writing custom code.

A high-performance, lightweight IDE built specifically for reviewing AI-generated code, featuring native Git/GitHub integration and seamless handoff to AI coding harnesses.