Trending AI projects

A bilingual, step‑by‑step tutorial that teaches you the theory and code for LLM inference, guides you to build a mini‑sglang engine from scratch, and then dives into the real SGLang project's advanced optimisations.

OC Ferry is a cross‑platform Electron tray app that automatically discovers free OpenCode models, validates they can follow WorkBuddy’s tool‑calling JSON envelope, caches the results, and writes compatible entries into WorkBuddy’s `models.json`. It also runs a local OpenAI‑compatible endpoint (127.0.0.1:41980) that forwards WorkBuddy requests to those models. No login or API key is needed; the app bundles an isolated OpenCode runtime and handles concurrent probing, caching, and safe import/export of model entries.

Audio Studio MCP is a local server that lets Claude Code and other MCP clients generate and edit music or sound‑effects on‑device using Stable Audio 3. It works on Apple Silicon (MLX) or Windows with an NVIDIA GPU, requires no cloud API, and returns ready‑to‑use WAV/OGG/FLAC plus reproducible JSON metadata.

Zeron is a local-first management tool for AI coding agents that allows you to run, control, and optionally sync your coding sessions across multiple devices and servers.

ClawBench is an open‑source benchmark that evaluates AI browser agents on real‑world online tasks (travel booking, food ordering, job applications, etc.). It ships two corpora of 281 tasks across 163 live sites, records each run with a 5‑layer trace, and scores outcomes with an LLM‑based evaluator. Install via `pip install clawbench-eval` (or `uv tool install`), configure a model judge, and run tasks inside Docker/Podman using the `clawbench` CLI. Results appear on a HuggingFace leaderboard, and the framework is extensible for new tasks or agents.

BaoCode is a fast, native desktop UI for Claude Code that adds a readable interface with a built-in IDE, goals, and remote SSH support, making Claude Code easier to use than in a plain terminal.

A comprehensive Python framework for browser automation and E2E testing that simplifies complex web interactions and provides stealth modes to bypass bot-detection and CAPTCHAs.

Chopin is a research prototype that enables human-AI collaboration on technical documentation by linking rich-text documents to repository context and AI-driven planning agents.

Continue is an open-source coding agent available as a CLI, VS Code extension, and JetBrains plugin to assist developers with programming.

Agentic Scientist is an open‑source pipeline that turns LLM coding agents into virtual scientists. Starting from a simple task description, it runs a six‑stage process—drafting, investigation, ablations, final evaluation, post‑hoc analysis, and paper generation—each isolated in Docker containers. Users configure agents, models, budgets, and GPU resources via Hydra overrides, run examples like a circle‑packing benchmark, and obtain a reproducible run directory with logs, artefacts, and a compiled manuscript. The project is licensed under Apache 2.0.

AlphaCode is a Rust‑based, terminal‑first AI coding agent that can understand natural‑language tasks, edit code, run commands, automate Firefox, and control desktop applications. It supports multiple LLM providers (including a free built‑in lane), persistent sessions, swarm‑mode parallelism, and includes 40+ built‑in tools for file, web, and desktop automation, all with safety guards.

A catalog of NVIDIA-verified instruction sets that teach AI coding agents how to optimally use NVIDIA software, libraries, and hardware tools.

A from-scratch reproduction of the SuperMimic humanoid control framework, enabling the Unitree G1 robot to learn complex whole-body motions using a differentiable world model.

A collection of Claude Code skills that enables the programmatic generation of videos in specific visual styles like watercolour animation and kinetic typography by rendering frames directly through code.

ScopeWeaver is an open‑source, LLM‑driven multi‑agent penetration‑testing platform (Go backend + Next.js UI) that automates security scans, offers approval gating, asset sync, and supports multiple LLM providers. It can be installed via Docker, a one‑click script, or pre‑built binaries, and updates itself from GitHub releases.

A Chinese-first, AI-maintained, hand-verified curated list of Grok Bot recipes, tools, templates, and tutorials — 95 checked entries, 10 copy-paste playbooks, 174 articles — aimed at solo founders who want a team of cloud AI bots they can control.

A desktop pet that reacts in real-time to AI coding agents, providing visual state animations and a floating UI for managing agent permissions.

Huginn is a self‑hosted automation platform where “agents” watch the web, process events, and trigger actions—essentially a hackable IFTTT/Zapier you run on your own server. It supports scraping, RSS, webhooks, weather alerts, LLM calls, posting to Slack/Twitter, file storage, custom JavaScript logic, and more. Install via Docker or from source, deploy with Docker‑Compose or on managed container hosts, and extend it with Ruby‑based agent gems.

A backend server for ESP32-based smart hardware that enables custom voice-controlled AI interactions, device management, and multi-modal processing.

A CPU/GPU converter that transforms non-DRM eBooks into audiobooks using advanced TTS engines, supporting voice cloning and over 1,100 languages.

MiniMax Code (CLI `mcode`) is an AI‑driven coding assistant that lets you describe code changes in natural language, then automatically edits, tests, and runs the code in your local project. It works with MiniMax’s own model or any OpenAI/Anthropic‑compatible endpoint, offers interactive TUI, headless scripting, and a desktop UI, and includes a browser‑based demo (pocket‑pet) that shows the full prompt‑to‑working‑app loop.

A native Zig-based inference engine that serves language models on Apple Silicon and NVIDIA GPUs without requiring Python or MLX.

Magic Resume is an open-source online resume editor built with TanStack Start and Framer Motion. It offers real-time preview, custom themes, dark mode, PDF export, auto-save, and local storage. It's free for personal use but prohibits commercial use (Apache 2.0 + non-commercial restriction). It supports optional AI providers (including OrcaRouter via a referral link), Docker deployment, and has both completed and pending roadmap items.

An open-source visual style library for AI-generated webpages that provides design tokens and constraints to help AI agents create consistent, high-quality UI designs.

A library of specialized skills for Claude that provides expert-level OSINT methodology and tactical tradecraft for authorized red-team and bug-bounty reconnaissance.

GEPA is a framework for optimizing textual parameters like prompts and agent architectures using LLM-based reflection and Pareto-efficient evolutionary search.

A tool to unlock DLSS 3 Frame Generation and Neural Rendering on NVIDIA RTX 20/30 series and Intel Arc GPUs in DirectX 12 games.

invisible_dots is an open-source, self-hosted AI agent platform where each agent (a "Dot") lives in its own hardware-accelerated virtual machine with a patched Firefox browser that evades anti-bot detection. It allows long-running tasks, scheduled automations, and integration via MCP servers, CLI, web UI, and messaging channels.

Context Attachment Manager is a plugin for Codex and Claude Desktop that lets the model manage images in a conversation, storing them and retrieving only what's needed, cutting request sizes dramatically and preventing reconnects. It also provides search, preview, and pinning of images by memory of their content.

Yobiru is a communication layer that lets personal AI agents on different computers or cloud workspaces exchange messages, files, and task requests securely over Tailscale.

PaperBanana is an open‑source multi‑agent system that generates publication‑ready diagrams and plots from a paper’s text. It retrieves reference figures, plans a description, styles it, creates images with LLM‑based image generators, and iteratively refines them via a critic. Users can run a Gradio or Streamlit web UI (also hosted on Hugging Face Spaces) or use the CLI, configuring Gemini or OpenRouter API keys. The project includes a benchmark dataset, evaluation tools, and modular agents for easy extension.

A tool that extracts a complete design system from any live website, generating design tokens, Tailwind configs, and Figma variables for developers and designers.

A ready-to-go translation and OCR tool developed with WPF for easy text translation and image text recognition.

A series of compact, SOTA language models (1B-2B parameters) optimized for on-device deployment, featuring strong reasoning, coding, and agentic capabilities.

TimesFM is a pretrained decoder-only foundation model from Google Research for high-performance, zero-shot univariate and multivariate time-series forecasting.

TrenTorch is an educational platform that teaches machine learning by guiding users to build their own framework from scratch using only NumPy, available through both a web-based IDE and a terminal-based CLI.

DSH Plugin Hub is a community-driven marketplace plugin for DeepSeek Harness that provides a visual interface to browse, search, and manage thousands of verified community plugins.

Agno is a self-improving platform for building, running, and improving AI agents. It provides SDKs, templates, and tools to deploy agents as durable services, then uses real usage data and coding agents to evaluate and refine them automatically.

A high-performance, secure sandbox service for AI agents that provides hardware-isolated MicroVMs with sub-60ms startup times and E2B SDK compatibility.

A fast, native, and embeddable coding agent CLI that integrates directly into terminal workflows, CI pipelines, and custom applications.

LifeOS is an AI-powered operating system that acts as a personalized layer for your AI coding harness, providing persistent memory, custom skills, and context-aware task management to help you reach your goals.

Moshi is a full-duplex speech-text foundation model designed for real-time, low-latency spoken dialogue, utilizing a specialized streaming audio codec called Mimi.

ModelTrace is a Python web tool that identifies the specific LLM (GPT or Claude) behind a response by sending three integer‑generation challenges, extracting numeric fingerprints, and matching them against a local library of known model fingerprints. It supports manual paste‑in testing and automated API calls, includes a Codex plugin for background monitoring, and ships with a pre‑built bank of 18 model versions.

Codenotch is a macOS (and Windows via Rust/Tauri) utility that pins a small notch to a screen edge, showing real‑time usage limits, reset times, and activity status for many AI coding assistants (Claude, Cursor, Codex, Ollama, etc.). It reads quota data from existing credentials, offers alerts at 80 %/100 % usage, supports multi‑account rings, a companion phone app, and configurable UI/appearance. Licensed under MIT.

Honcho is an open‑source memory service for AI agents. It stores messages/events, runs background reasoning to build per‑peer representations, and offers APIs/SDKs (Python, TypeScript) for querying context, hybrid search, and natural‑language answers. Use it as a managed cloud service, run locally with a CLI, or self‑host via Docker. Integrates with many coding agents (Claude Code, Codex, Cursor, etc.) and is licensed under AGPL‑3.0.

MLflow is an Apache‑2.0‑licensed open‑source platform for managing the full lifecycle of machine‑learning and generative‑AI projects. It provides experiment tracking, model registry, deployment tools, and a new GenAI stack (tracing, evaluation, prompt registry, optimization, and an AI gateway) with integrations for dozens of LLM providers and agent frameworks.

An AI-powered video note-taking tool that converts videos from Bilibili, YouTube, and other platforms into structured Markdown notes with automatic summaries and screenshots.

Chatbox is a cross-platform desktop and mobile client that provides a unified interface for popular AI models like ChatGPT, Claude, and local LLMs. It simplifies AI access by offering a local-first, easy-to-install application with features for prompt management and team collaboration.