Trending AI projects

headcount is an open‑source library that packages AI agent skills into company‑style departments (e.g., Security, Marketing, Finance). Each department is a plug‑in containing named skills that load on‑demand, letting you build modular, specialist‑driven LLM workflows. Install via Claude Code or ChatGPT, then ask questions or invoke `/department:skill` directly. The repo ships 16 departments, 172 skills, and an interactive org‑chart for browsing.

A generative, code-rendered music video system that uses audio analysis and three.js to create deterministic, word-synced visuals exportable in 4K60.

An AI-powered English listening and speaking training app that automates a scientific loop of intensive listening, shadowing, and retelling with built-in spaced repetition.

A free, locally-run Google Maps scraper for lead generation, packaged with a Claude Code skill so AI coding agents can drive scrapes in plain English and deliver clean CSVs of business contacts.

A self-hosted, privacy-focused alternative to NotebookLM that lets you organize multi-modal research, chat with your documents, and generate podcasts using your choice of AI models.

Chrome DevTools for agents (`chrome-devtools-mcp`) is an npm‑packaged MCP server and CLI that lets AI coding assistants control a live Chrome browser via DevTools. It provides performance tracing, debugging, automation (via puppeteer), and a set of pre‑written “skills” for agents like Claude, Copilot, and Gemini. Install with npm, configure the MCP client, and agents can issue natural‑language prompts (e.g., “check the performance of a site”) to launch Chrome, run audits, and receive actionable insights.

9Router is a Node.js/Next.js service that acts as a local OpenAI‑compatible gateway for AI coding tools. It compresses payloads (RTK Token Saver), tracks quota, and automatically falls back across paid, cheap, and free LLM providers, supporting 40+ APIs and self‑hosted servers. The goal is to keep coding uninterrupted while cutting token costs by 20‑40 %.

helm-d is a DeepSeek Harness plugin that enforces evidence-based security analysis—every conclusion must cite disk-based E-numbered evidence, and it ships 34 tools, 362 on-demand references, and hot-swappable jailbreak profiles for Android, Web, Native, Protocol, Malware, and AI-Security tasks.

PandaCode is an open‑source native desktop app that lets you run many AI coding agents (Claude Code, Codex, Kimi Code, etc.) alongside Git, a terminal and a browser. It offers a Kanban task board, agent‑team orchestration, bot integrations, local/cloud sandboxes, one‑click deploys and deep Git support, all stored locally with no telemetry.

An ultra-lightweight, self-hosted personal AI agent framework that integrates multiple LLM providers, long-term memory, and various chat apps into a single runtime.

Monid is a Deno‑based platform that aggregates thousands of third‑party tool APIs (search, enrichment, media generation, etc.) behind a single, agent‑friendly endpoint. Developers write declarative connectors describing a provider’s endpoints, auth, and usage model; the engine discovers the best endpoint, executes the call, and bills per the raw response.

AI破甲 is a Codex configuration and Skill installer that sets up a custom instruction file and Skill to guide AI in authorized tasks, with verification and recovery tools. It is based on a modified version of gpt-instruct v45, but explicitly removes certain aggressive behaviors. It does not claim universal compatibility or proven demo success.

Dioramas is a free, open-source framework for building cinematic, interactive 3D websites. It uses AI to generate and optimize 3D hero models from text prompts, and includes a small three.js engine for visual effects, plus 20 example sites.

A bridge that integrates ChatGPT Web models into the Codex workflow, allowing users to use their ChatGPT account plans and tools via browser automation and MCP.

RP1 is an open-source full-stack platform for building and controlling humanoid robots, covering everything from mechanical design and electronics to firmware, control, and AI learning. It gives developers the complete stack to reproduce, understand, and extend a human-sized humanoid.

Claude Switcher is a native macOS (and beta Windows/Linux) app that audits 14 network/system fingerprints before launching Claude AI. If any check fails, the app refuses to start, and a background kill‑switch instantly terminates Claude on tunnel loss. It includes a hand‑crafted STUN probe for WebRTC leaks, per‑account snapshot isolation, and self‑healing configuration.

This project is an independent reproduction of the SuperMimic method for whole-body humanoid control. It trains a policy to track LAFAN1 motions using a differentiable world model, achieving results on par with the paper in under 1.5 GPU-hours on a single GPU.

Goutoujunshi is an open‑source AI skill for conversational assistants that analyses a user’s emotions, relationship context and chat logs, then provides evidence‑based, actionable dating or relationship advice. It includes a curated 135‑item knowledge base, optional long‑term memory (with consent), and safety checks, all packaged under an MIT license.

TextFold is a local desktop/CLI app that uses your own LLM to create five hierarchical “folds” of any text (papers, web pages, PDFs, Word docs, etc.), letting you zoom from full paragraphs down to a one‑line topic map. It integrates with Zotero, supports 12 languages, works on macOS/Windows/Linux, and provides plugins for Claude Code and Pi so LLM replies can be folded the same way.

A marketplace of Markdown‑defined plugins, agents, skills and commands that extend AI‑coding tools (Claude Code, Codex CLI, Cursor, etc.) with ready‑to‑use LLM‑driven workflows like scaffolding, testing and security scanning.

A collection of expert-level instructions and scripts that extend AI coding agents with specialized capabilities for web development, performance auditing, and documentation standards.

ste‑zh is an open‑source skill that makes LLM agents write their Chinese status reports in a strict, evidence‑based format inspired by the aerospace standard ASD‑STE100. It can be installed via the `skills` CLI or manually, activated with simple commands, and is useful for teams that need clear, concise AI‑generated updates.

A complete, free open-source course where you build your own coding AI assistant from scratch, learning the engineering layer (harness) that makes such assistants actually work — including sandboxing, context management, subagents, and evaluation — across 8 lessons with articles, videos, and runnable code.

A high-performance Rust-based browser automation CLI designed to give AI agents reliable, semantic control over web pages and browser state.

A desktop workspace that coordinates multiple AI coding agents by providing a unified Kanban interface, isolated execution environments, and a project-level orchestrator for planning and delegation.

WeVisOne-VL is a Tencent vision-language model that unifies object-level tasks (detection, counting, pointing, OCR, referring) and pixel-level tasks (segmentation, matting) in one system, using a 9.6M-parameter pixel decoder and structured text tokens. It comes in 2B and 4B sizes, reliably rejects absent targets, and outperforms open baselines across all headline benchmarks. Inference code, demos, and evaluation pipelines are released; fine-tuning code is still pending.

A comprehensive guide and technology stack for building, launching, and scaling profitable AI applications while minimizing infrastructure costs.

Mem0 is an open‑source memory layer for LLM‑based assistants, providing multi‑level, hybrid retrieval (semantic, BM25, entity linking) and temporal reasoning. It offers a Python/JS SDK, CLI, self‑hosted Docker stack, and a managed cloud service. New 2026 algorithm improves token efficiency and latency, with benchmark scores up to 94.4 on LongMemEval. Ideal for building personalized chatbots, support agents, or any AI system that needs long‑term, user‑specific memory.

A framework that decomposes single anime illustrations into up to 23 semantically distinct, inpainted layers exported as a PSD for 2.5D animation and rigging.

This explainer covers the purpose, content, prerequisites, technologies, languages, and community aspects of the AI Agents for Beginners course, all grounded in the README.

A bridge plugin that enables AI agents to directly control Blender via a TCP channel, allowing for automated scene editing, viewport monitoring, and complex 3D optimization tasks.

YuE2 Studio is a standalone Windows application for generating, editing, and training AI music models, providing a complete workflow for creating songs from lyrics and scores without needing Python.

Kitten TTS is an open-source text-to-speech library that supports real-time voice cloning, multilingual synthesis, and expressive speech control, designed to run efficiently on both CPUs and GPUs.

A research‑grade code base that reproduces *Internal Safety Collapse* – a failure mode where a coding‑agent, while iteratively editing a task workspace, ends up generating unsafe content it would normally refuse. The repo provides Dockerized TVD (Task‑Validator‑Data) workspaces, examples with LlamaGuard, Detoxify, Qwen3Guard, and a large‑scale audit of 62 frontier LLMs, plus guides to run, inspect, and extend the experiments.

TraceRoot is an open‑source platform that records AI agent calls, flags bad behaviour with custom detectors, clusters issues into signals, and runs automated evals so a coding‑agent can suggest and verify fixes, closing a self‑improvement loop. It offers a CLI, Python/TS SDKs, and integrations with major agent frameworks and model providers, and can be used via hosted cloud or self‑hosted Docker Compose.

A high-performance browser agent that uses a dynamic, indexed action space and structured DOM state instead of screenshots to achieve ultrafast web automation.

WeChatBridge is a native macOS (and Windows) tool that adds a WeChat Share Extension, letting you forward chat records directly to AI agents like Codex, Claude, or Doubao, or archive them into Obsidian as Markdown — all processed locally with privacy in mind.

Lyapunov is an LLM-first desktop framework for robotics and 3D simulation that integrates scene construction, robot preparation, and physics engine control into a single workspace.

TradingAgents‑CN Community Edition is a hybrid‑licensed platform that uses multiple LLM‑driven agents to research Chinese A‑share stocks. It offers debate‑style single‑stock analysis, natural‑language screening, a paper‑trading sandbox, and free learning courses. Built with FastAPI, Vue 3, MongoDB, and Redis, the core engine is Apache‑2.0 while the application layer is source‑available for non‑commercial use.

A library of over 2,600 installable playbooks that allow AI agents to search, select, and apply reusable instructions for complex technical tasks.

A platform that provides AI coding agents with up-to-date, version-specific library documentation and code examples to prevent hallucinations and outdated code generation.

A personal agent application with a persistent browser, Linux terminal, and Google integrations that allows users to delegate complex tasks and review agent actions.

Official code for the book *Build a Large Language Model (From Scratch)* – a step-by-step, from-scratch implementation of a GPT-like LLM in PyTorch, covering text preprocessing, attention, model implementation, pretraining, and finetuning. Runs on laptops, with optional GPU acceleration.

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).

The project is a curated, tested list of open-source video-making skills for coding agents, with categories, selection criteria, and end-to-end test results.

Hyperresearch is a deep research harness for AI coding agents (Claude Code, OpenAI Codex) that runs a 16-step pipeline to gather hundreds of sources, verify citations, and produce adversarial-audited reports with a persistent searchable vault.

A human-curated collection of 1,497+ Agent Skills (instruction bundles for AI coding assistants) from official teams like Anthropic, Google, Stripe, and Microsoft, plus community contributions. Compatible with Claude Code, Cursor, Copilot, and other tools — focused on real-world use by engineering teams, not AI-slop.

Roast Room is a Node.js web app that uses Claude Code sub‑agents to simulate a council of seven famous business figures. You feed a decision, the agents argue, score options, and produce a verdict PDF—all driven by curated “brains” built from each person’s public work.