Trending AI projects

Mobilerun is an open-source framework that enables AI agents to control Android and iOS devices, allowing for automated mobile workflows through natural language commands.

Webnovel Writer is a Claude Code plugin that lets authors write long‑form web novels with AI while preserving plot consistency. It offers slash‑commands for initializing a book, planning outlines, writing and reviewing chapters, querying stored facts, and viewing a read‑only dashboard. Facts are recorded in a “Story System”, indexed with RAG (or BM25 fallback), and projected into summaries and long‑term memory so the model remembers characters, foreshadowing, and world rules across hundreds of chapters.

EAC design world is a local video/design studio plugin for DeepSeek Harness that lets AI and humans collaborate on creating videos, websites, slides, documents, and images with real files, previews, and exports. It includes a design library, motion shots, recording, editing, color grading, subtitles, and more, and can run standalone or inside Harness. It requires Node.js 22+, FFmpeg 7+, and optionally Chrome/Edge; it’s AGPL-3.0 licensed.

Happy is an MIT‑licensed, multi‑provider LLM harness that lets you run Claude, Codex, and Grok from a single CLI, mobile app, or desktop client. It syncs sessions across devices with end‑to‑end encryption, supports real‑time multiplayer collaboration, and reuses your existing subscriptions. The repository contains the mobile/web client, original CLI, and the encrypted relay server; companion repos provide the desktop client and the open‑source agent runtime.

DSH Voice AI Girlfriend is a plugin that turns DeepSeek Harness into a talking companion: it listens through your microphone, responds with a cloned voice, shows a lip-synced digital human avatar, and can push replies to QQ.

GEO 医生 is an open‑source skill/plugin for AI assistants (Claude Code, WorkBuddy, etc.) that audits a website’s AI‑crawler accessibility, extracts verified product facts, generates realistic user questions, samples answers from major LLMs, analyses recommendation gaps, auto‑writes and publishes SEO‑friendly content, and can retest weekly via GitHub Actions. Install with a single chat command or run a one‑line Node.js audit; the tool is MIT‑licensed and focuses on improving AI‑search (GEO) visibility for Chinese and global products.

An open-source evaluation framework for AI applications and agents that provides a library of ready-to-use graders and tools to generate custom evaluation rubrics.

Blender Agent Skills is a set of 22 scripts that let AI coding assistants (Claude, Codex, Gemini, Copilot, etc.) control Blender 5.2+ for full‑pipeline 3D asset creation, from modeling to export and QA. Install via a Claude plugin or by cloning and running the provided installer; the skills are then callable by the supported agents.

Scenario for Blender is a GPL‑licensed Blender add‑on that connects to Scenario’s cloud AI services. It lets users generate images, video, 3‑D meshes, PBR materials, audio, 360° environments and block‑out layouts from text prompts or reference assets, then import the results back into the current Blender scene. The plugin includes UI panels, a local MCP server for agent integration, and requires a Scenario account (credits are consumed per generation).

GameScaling is a set of vision-language models that were continued pre-training on 1,600 hours of keyboard-and-mouse gameplay so they can play computer games in real time. Each model takes the current screen (plus a short history) and outputs the next 200 ms of input as six 33 ms key steps plus one mouse move and scroll; the repo provides the inference client, system prompt template, examples, and tests for serving the released checkpoints.

iPolloWork is a local‑first, Electron‑based workbench that unifies multiple AI‑agent runtimes (OpenCode, Codex, DeepSeek Harness) into one collaborative UI for teams. It lets you manage agents, plugins, tasks and editable outputs (code, docs, designs, video) from a single workspace, with optional cloud integration for identity and hosted workers. Installers are provided for macOS, Windows and Linux; the source can be built with Node 22+, pnpm and Bun. The project is released under a source‑available license that permits personal use but requires permission for larger or commercial deployments.

Browser UI for the self-hosted Hermes autonomous agent, with chat, sessions, file browsing, voice input, profiles, and security — all parity with the CLI, no build step.

A set of Claude Code/Gemini‑CLI plugins that turn design knowledge into machine‑readable skills and ready‑to‑run commands, letting AI agents execute full design workflows (research, strategy, UI, interaction, systems, ops, critique) directly from the assistant.

A code-based art and music engine that generates deterministic, hand-drawn style illustrations and animations, eliminating the unpredictability of traditional AI image generation.

oMLX is a macOS-native LLM inference server that optimizes local model performance through tiered KV caching and continuous batching, managed via a menu bar app and web dashboard.

EverOS is a Python library that provides local‑first, Markdown‑backed memory for AI agents. It syncs files to SQLite and LanceDB, offers a simple HTTP API for add/flush/search, and works with a single OpenRouter key for keyword search. Optional extras add embedding‑based hybrid search, reranking, and multimodal file ingestion.

MemOS 2.0 is a Memory Operating System for LLM‑based AI agents. It offers a unified API to store, retrieve, edit, and delete multi‑modal memories (text, images, tool traces) organized as a graph. The system can run as a hosted cloud service (API‑key access) or be self‑hosted via Docker (Neo4j + Qdrant) or a local SQLite plugin. Ready‑made plugins integrate MemOS with OpenClaw, Hermes, and DeepSeek Harness. Benchmarks show it outperforms existing commercial memory products. Licensed under Apache 2.0.

A Microsoft-produced, project-based 10-week introductory data science course with 20 lessons, quizzes, hands-on projects, and 50+ language translations — free and open source.

RikkaHub is a native Android chat app that lets you talk to many LLM services (OpenAI, Claude, Gemini, etc.) from your phone. It supports multimodal inputs, image/video generation, custom provider configurations, and simple agent‑style extensions, all wrapped in a Material‑3 UI.

A collection of structured instruction sets for AI agents that provide specialized workflows for goal setting, disk cleanup, research, and writing, following the Agent Skills open standard.

An open-source, AI-first call center platform that integrates with FreeSWITCH to provide seamless transitions between AI voice agents and human operators within a single unified record.

Sweety‑handdraw‑video is an open‑source skill that turns a topic, article or story into a hand‑drawn narrated video. An LLM‑based Agent writes the script, designs the storyboard and picks a style; local Python tools then draw each frame, render handwritten subtitles, sync with a locally‑generated voice‑over and export a 1920×1080 MP4 via FFmpeg. It includes 21 illustration styles, six hand‑arm presets, six built‑in voice profiles, and runs entirely on the user’s machine (macOS/Linux verified).

A native Mac AI study and research assistant that integrates document reading, AI-powered note-taking, and task management into a unified workspace.

A Chinese‑first AI‑assisted editing skill that removes boiler‑plate and stiff phrasing while keeping facts, conditions and the author’s voice. Load the three markdown rule files into a chat model (Claude, Grok, ChatGPT, etc.) and ask it to rewrite or just flag a passage. Provides mini, lite and full installations, granular control over how much restructuring is allowed, and a benchmark of 123 cases showing zero hard failures.

A vLLM fork that serves Jev-style decision LLMs — small models that pick a label and probability from candidate answers — with structured Choice/Noul/Score APIs, multimodal (text/image/video) support, and backends for CUDA, Ascend, Apple Silicon, and experimental Windows CPU.

A local, privacy-focused system that turns real-time bird audio detection into a beautiful, automatically updating digital art frame using historical natural history illustrations.

An open-source, local-first 3D building editor built with WebGPU that allows users and AI agents to model buildings and manage facility data.

code‑review‑graph is a Python package that parses a repository with Tree‑sitter, stores the AST‑derived nodes/edges in a SQLite graph, and serves a minimal, blast‑radius slice of that graph to any MCP‑compatible AI coding assistant. It reduces per‑question token usage by ~63×, supports 30+ languages (including notebooks), offers incremental updates, framework‑aware edges for PHP/Laravel and Java/Spring, a GitHub Action for CI risk‑scored reviews, and optional semantic embeddings. Benchmarks show large token savings and an average F1 of 0.69 for impact analysis. The tool is aimed at developers and teams using LLM‑based code reviewers who need precise, low‑overhead context.

A lightweight, provider-agnostic framework for building multi-agent workflows, supporting text, voice, and sandbox-based agents.

A typed Python AI SDK for building extensible agents with structured outputs, durable execution, and support for multiple model providers and interfaces.

An enterprise world model and bitemporal knowledge graph that tracks how information evolves over time to provide a trustworthy, auditable foundation for AI agents.

WhisperX is a Python library that speeds up OpenAI’s Whisper transcription by batching, adds per‑word timestamps via forced alignment with wav2vec‑2.0, and can label speakers using pyannote‑audio. It runs on GPU (or CPU), supports several languages, and is installable via pip.

Comp AI CRM is an open-source CRM where an autonomous AI agent runs the show — reading emails, researching contacts, and recording only observed facts. It’s built with modern tooling (Next.js, NestJS, Postgres), can connect to external services (Gmail, LinkedIn) with granular permissions, and runs with or without external API keys. The agent never guesses, operates in a safe sandbox, and lets you watch and talk to it on every record. Quick start requires Bun, Docker, and a few environment variables.

An MCP server + Ghidra plugin that lets LLMs autonomously reverse-engineer binaries — decompiling code, listing symbols, and renaming methods through standard MCP clients like Claude Code or VS Code.

DataFlow is an open‑source Python framework that lets you build, run, and share low‑code pipelines for generating, cleaning, and evaluating LLM training data. It provides a library of reusable operators, a visual WebUI, an AI‑assistant for auto‑creating pipelines, and a Ray‑based distributed execution layer.

OpenObserve is an open-source, cloud-native observability platform that combines logs, metrics, traces, RUM, and AI monitoring in a single tool, claiming up to 140x lower storage costs than Elasticsearch via Parquet columnar storage and S3-native design. It offers a cost-effective, self-hostable alternative to Datadog, Splunk, and Elasticsearch with a single-binary deployment and OpenTelemetry native support.

A repository of reusable instruction and script packages that AI agents can use to perform specific tasks in a repeatable way.

A Korean-specific AI writing tool that removes machine-generated stylistic markers like translation-style phrasing and repetitive structures while preserving original facts and meaning.

A local-first data workspace that connects multiple company data sources into a unified ontology for private, AI-powered exploration and querying.

Screenpipe is a local-first, source-available application that continuously records your computer activity to create a searchable AI memory, enabling users to automate workflows and provide context to AI agents.

A fast and local neural text-to-speech engine that provides offline speech synthesis using espeak-ng for phonemization.

Open‑source AI‑driven video/short‑film creation platform with a visual canvas, workflow management, and async generation of text, images, video and audio. Built with React (frontend) and Go/Gin (backend), deployable via Docker, and extensible through a skill ecosystem.

A curated catalog of 258 agent-native services and infrastructure across 16 categories, designed for AI agents to discover, evaluate, and connect to email, browsers, memory, payments, sandboxes, and more directly via MCP, CLI, or SDK onboarding.

Droidline is an automation tool that lets you control Android phones from your PC using code or AI agents without needing ADB, USB cables, or root access.

Pier is a macOS terminal manager that centralizes multiple AI coding agent sessions, allowing developers to monitor, edit files, and manage Git workflows in one place without changing their existing command-line tools.

Locals is a local AI model server for Macs built for coding agents, not plain chat. It runs models on your own GPU, batches requests from many agents at once, and enforces valid tool calls — delivering up to 3x faster answers for parallel agents and long conversations, while keeping all data on your machine.

A collection of 42+ runnable Jupyter notebooks teaching Retrieval-Augmented Generation techniques, from basics to advanced, with code, intuition, and references.

cinetic is an agent skill that lets coding agents create cinematic launch films, product videos, and logo animations from a short text brief, handling concept, branding, sound, and final render with professional quality.