Computer vision and AI

  • danielgatis/rembg (Python, 25347 stars): removes the background from images and can be used from the command line, as a library, as an HTTP server or in a Docker container. It handles whole folders in batch and offers CPU, CUDA and ROCm backends, plus alpha matting options to refine the edges.
  • lucidrains/vit-pytorch (Python, 25561 stars): a PyTorch implementation of the Vision Transformer, which classifies images by applying a single transformer encoder to image patches. It collects variants such as Simple ViT, NaViT, CaiT, MaxViT and Dino, plus masked image modeling models.
  • jtydhr88/ComfyArt (TypeScript, 457 stars): an open source design canvas that runs locally on top of ComfyUI, with an agent that works on the canvas through tools. Images, video, audio and 3D models are generated on your own hardware, with the option to switch to Comfy Cloud without downloading models.

Platforms and productivity

  • nocobase/nocobase (TypeScript, 24840 stars): a platform for building business systems in which AI agents write the code, while permissions, workflows and approvals are already in place. It includes a core application and a governance studio to supervise the agents' work, and starts with Node.js 24+ and pnpm 11, using SQLite with no separate database server.
  • macro-inc/macro (Rust, 6166 stars): a single workspace bringing together email, chat, documents, tasks, calls and CRM. Cross-references between tools are stored as a bidirectional graph and the team shares an AI memory; the interface is in SolidJS and the backend in Rust.
  • nextai-translator/nextai-translator (TypeScript, 25027 stars): a cross-platform browser extension and desktop app for translating, reviewing and summarizing text through OpenAI-compatible LLM APIs. It includes streaming translation, screenshots, text-to-speech and a vocabulary book with reminders generated from saved words.

Terminal and system tools

  • fastfetch-cli/fastfetch (C, 25281 stars): a command-line tool that displays system information with a visual layout similar to neofetch. Written mostly in C for high performance, it supports Linux, macOS, Windows 8.1+, Android, several BSDs, Haiku and SunOS, and the displayed modules are configurable through JSONC files.
  • charmbracelet/gum (Go, 24567 stars): brings interactive components such as selections, inputs, filters, spinners and tables into shell scripts without writing Go code. Built on Bubbles and Lip Gloss, it returns results on stdout, so they can be saved into variables or files.
  • opennookorg/betterwispr (Swift, 508 stars): transcribes speech into text on macOS and types it into the foreground app while a key is held down. Recognition runs on the Mac with Whisper, Parakeet or Apple's own engine, without sending audio to external servers unless explicitly chosen, and it learns from the user's corrections.

Automation and agents

  • browser-act/skills (Python, 6325 stars): a browser automation CLI designed for AI agents, which gets past anti-bot systems and hands control to a human operator when the agent gets stuck. It runs multiple tasks in parallel with isolated sessions and accounts, using stealth fingerprints, TLS rotation, residential proxies and CAPTCHA solving.
  • KnifeLemon/Droidline (Go, 459 stars): lets you automate real Android phones from Python, Node.js, the CLI, HTTP or Appium clients, without ADB, a USB cable or root. You install a single app on the phone, enable accessibility permissions and pair the two with a QR code; the phone connects outward to the PC, so it also works on the move through a tunnel, with end-to-end encryption.
  • reacher-z/Awesome-CUA (Python, 490 stars): an open, verified catalogue of computer-use agents, with papers, GUI models, datasets, benchmarks and tools exportable to JSONL and SQLite. It publishes sources, review status and open gaps, separating bibliographic verification from relevance assessment.
  • AIshihzong/awesome-reverse-engineer-anything (Python, 489 stars): a curated list of projects in which AI agents took apart and rebuilt closed software, stating whether REA or other tools were actually used. Each entry is marked as verified, AI-RE, rebuild or to be confirmed, with the verification method declared.

Gaming and development

  • Carter54git/Greates-Games-Decompilations-List (N/A, 456 stars): a list of games for which a public decompilation, disassembly or static recompilation exists, grouped by platform. Each entry names the project that actually did the reverse engineering; many works are incomplete and the repository hosts no game assets.