Trending AI projects

Open SEO MCP Skills is an MIT‑licensed set of Claude plugins that let Claude query your own Google Search Console, GA4, Google Ads, and DataForSEO data via a Ryze MCP server, enabling free, real‑time SEO audits, keyword research, rank tracking, competitor gap analysis, backlink checks, and AI‑visibility reports.

Folkbench is a benchmarking platform and desktop tool that ranks AI model providers and allows users to easily switch between them in various coding assistants without manual configuration.

Skill Recorder is a desktop app that records a single user task (screen, clicks, terminal, optional narration) and uses GitHub Copilot to turn it into a reusable `SKILL.md` procedure or scheduled automation for AI agents.

Anthropic Sandbox Runtime (srt) is a cross‑platform tool that uses native OS sandboxing (sandbox‑exec, bubblewrap, Windows Filtering Platform) to run arbitrary commands with strict filesystem and network restrictions. Designed for AI agents and MCP servers, it offers a CLI (`srt`) and a TypeScript library, supports dynamic config updates, per‑command allow lists, and real‑time violation reporting.

DSH Desktop is a native Windows/macOS client that packages the DeepSeek Harness AI agent platform into an easy‑to‑install desktop app, offering a built‑in plugin market, remote mobile control, and configurable LAN access.

A community‑curated library of structured prompts for the REA AI reverse‑engineering agent, covering binaries, firmware, Android, games, kernels, and security research. Prompts enforce evidence‑backed, single‑feature investigations and guide users from installation through clean‑room cloning and vulnerability analysis.

Strix is an open-source autonomous AI penetration testing platform that scans code and applications to find, validate, and help fix security vulnerabilities.

Ruflo is an orchestration harness for Claude Code and Codex that adds memory, agent swarms, and security controls to enable collaborative, self-improving AI workflows.

Claude Code Game Studios is a template that adds a full AI‑driven studio hierarchy (49 agents, 74 slash‑command skills, 15 safety hooks) to a Claude Code session, letting you manage game design, coding, art, QA, and release with configurable rigor levels.

AutoSwarm‑Shadow‑Adapter is a plug‑in for the AutoSwarm AI‑agent orchestration framework that provides a non‑blocking human‑in‑the‑loop (HITL) and BYOC (bring‑your‑own‑compute) layer. It intercepts tasks needing human or remote GPU review, parks them in an isolated SQLite WAL database, and returns an `AWAITING_HUMAN` state so the rest of the DAG continues. A FastAPI‑backed REST API and React 19 UI let humans claim tasks, view markdown briefs, and upload up to 5 files (≤ 50 MB). Remote workers run a `shadow‑worker` daemon, authenticate via JWT, execute a local model (Ollama, Anthropic, OpenAI, etc.) and push results back. The adapter writes an approval phase into AutoSwarm’s store DB, automatically resuming downstream nodes. Installation is a single `pip install`, then register the adapter in AutoSwarm’s registry and launch the portal (or side‑car server). Configurable via environment variables, it offers audit‑ready logging, zero core code changes, and a centralized artifact registry. Ideal for solo founders, freelancers, or regulated enterprises needing asynchronous human sign‑off without stalling parallel AI work.

colibri is a lightweight C engine that runs huge mixture-of-experts AI models on ordinary CPUs by streaming rarely-used expert weights from disk, with optional GPU acceleration via Vulkan or CUDA. It supports a wide range of open models (language, vision, decision) and offers a one-step installer, a browser dashboard, and OpenAI/Anthropic-compatible APIs.

Caveman is a Node‑CLI that wraps LLM coding agents (Claude, Gemini, OpenAI, etc.) and reduces token usage by shortening model answers, compressing input files, auto‑selecting cheaper models, and caching repeated scripts. It claims 30‑+% fewer input tokens, 50 % lower cost, and no quality loss, backed by studies from Adobe, JetBrains, and Elastic.

CLI‑Anything is an open‑source Python framework that automatically generates agent‑friendly command‑line wrappers for any software. It ships a package manager (`cli‑anything‑hub`) that indexes dozens of community‑built CLIs (GIMP, Blender, Zoom, GIS tools, etc.) and lets agents discover, install, and invoke them via structured JSON output. Developers can use a 7‑phase generator to create new CLIs for any application, and AI agents (Claude Code, Cursor, Pi, OpenClaw, …) can install the meta‑skill to browse the registry autonomously. The project is actively maintained, well‑tested (2 461 passing tests), and Apache‑2.0 licensed.

Spec Kit is an open-source toolkit that provides structured, repeatable processes for AI coding agents to handle feature development, bug fixing, and project assessment.

Printfilm is a self-hosted, step-by-step AI production platform that helps users create short videos and manga-style dramas by managing the entire workflow from scriptwriting to final video synthesis.

Hypit is an open-source tool that lets AI agents clone or create videos as reusable workflows, with pluggable components and optional AI generation, costing only model-service fees.

text-to-cad is a plugin that lets AI assistants generate 3D models from text or images, create engineering drawings, check designs for manufacturability, and send parts to fabrication services. It works with popular agent apps like Claude Code, Codex, Cursor, Gemini, and Grok, running locally via its own CAD server.

Pi is a minimal, extensible agent harness that allows developers to build and customize their own AI coding assistants through modular extensions and skills.

A set of eleven Claude‑compatible “skills” plus six Python utilities that help YouTubers script, edit, optimise and audit videos—all run locally on plain Python 3 and never publish automatically.

A web UI plugin marketplace for DeepSeek Harness, enabling browsing, one‑click install, updates, theming, backup, diagnostics and secure management of thousands of community plugins.

A genealogy skill for AI agents that turns them into evidence-first family-history researchers for the former USSR, with structured registers, historical jurisdiction handling, name-variant support, audit scripts, and strict responsible-use guardrails.

OpenDesign is a desktop‑first, Apache‑2.0‑licensed workspace that lets teams use any installed LLM coding agent (Claude, Codex, DeepSeek, etc.) to generate and iterate on design artefacts—web prototypes, slide decks, dashboards, images, and motion‑graphics videos. It ships with hundreds of design‑system packages and template libraries, supports 26+ agent CLIs via a simple plug‑in system, runs locally (macOS/Windows, source‑build for Linux), and can export results as HTML, PDF, PPTX, or MP4 via the open‑source HyperFrames engine.

OpenFilm is an open-source tool that enables AI coding agents to generate and edit videos using standard web technologies, providing a browser-based studio for human refinement and professional rendering.

Gemini CLI is an open‑source Google tool that brings Gemini LLMs to the terminal, enabling code generation, debugging, and automation via an interactive REPL or scripted commands, with multiple auth methods and extensible MCP‑server integrations.

A library of 177 scientific and research skills that provide curated documentation and workflows for AI agents to interact with scientific databases, Python packages, and research methodologies.

CodeGraph builds a local, up‑to‑date knowledge graph of a codebase (30+ languages) and plugs into AI coding assistants (Claude, Cursor, Copilot, etc.). By providing agents with a single, precise payload of symbols and call‑paths, it cuts tool calls, token usage and latency—benchmarks show up to 44 % cost reduction and 88 % fewer tool calls.

DeepOpen is a non-autoregressive SDK and framework for structured decisions: it classifies text or JSON into predefined categories, returns scores/probabilities, and does not generate free-form text. It packages the upstream Laya model with routing, question templates, and reproducibility experiments, while clearly documenting limitations like overconfidence, low reliability on some tasks, and the need for validation on your own data.

An operating layer for Hermes Agent that adds structured workflow management, intelligent model routing, parallel task execution, and a persistent memory system to improve agent reliability and efficiency.

A multi-agent prediction engine that creates high-fidelity digital sandboxes to simulate and predict the outcomes of real-world events or fictional scenarios.

img2threejs is an open‑source skill that lets LLM agents turn a single reference image into a procedural Three.js model written in TypeScript. It uses a gated, Python‑driven pipeline to enumerate visual details, generate code‑only geometry, and produce animation‑ready assets. Installable via `npx img2threejs@latest`, it works with Claude Code, Codex, OpenCode, Hermes and supports optional plugins for characters, CS2 weapons, etc. The project includes a live demo gallery and is Apache‑2.0 licensed.

A desktop companion plugin for DeepSeek Harness that provides an interactive, animated pet capable of roaming the browser and desktop while reflecting AI agent status and account information.

Claude-Guardian is a "Skill" package for AI coding agents (like Codex) that guides users through safely setting up a Claude usage environment — covering route selection (local/VM/VPS), fixed exit IPs, account registration, firewalls, and network optimization. It uses a three-tier authorization system (A/B/C) so users only need to approve high-risk changes, includes verification and rollback scripts, and never handles payments, credentials, or Claude access directly.

OpenUI is a framework‑agnostic, streaming‑first UI library that lets LLMs generate UI components using a compact language (OpenUI Lang). It provides parsers, prompt generators, and runtime bindings for React, Vue, Svelte, Angular, plus a CLI for scaffolding apps. The language reduces token usage by ~50 % versus JSON, enabling progressive rendering of model‑generated interfaces.

VoxCPM is a tokenizer-free, diffusion-based Text-to-Speech system that provides studio-quality multilingual speech generation, creative voice design, and high-fidelity voice cloning.

dsh‑purge is an open‑source red‑team plugin for DeepSeek Harness (web and desktop). It adds a UI dock that lets you apply custom prompt‑injection patches, view patch status, and restore the original host policy. Install via `dsh plugin --profile <web|desktop> add dsh-purge`, then click **Apply** in the new dock. The tool is MIT‑licensed, requires attribution, and is intended only for local, authorized security testing.

A curated marketplace directory for Claude Code plugins, split into Anthropic-maintained and third-party sections, with strict naming rules, a standard plugin structure, and a special mechanism for bundling skills from repos that lack a plugin manifest.

本机优先的 DSH 插件，消除反复确认、提供状态自检与快照回滚，令 AI 助手更连贯地执行学习/实验任务（改交互层，不改模型）。

Switch is a self-hosted platform that integrates AI agents directly into your team's existing messaging apps, turning chat channels into collaborative workspaces where humans and agents share context and workflows.

A free, MIT‑licensed collection of markdown‑based “Agent Skills” that teach AI coding agents how to perform a wide range of marketing tasks—from SEO audits and ad‑creative generation to CRO and growth‑loop automation. Installable via a generic CLI, Claude Code, Codex, or SkillKit, the library works with any agent supporting the Agent Skills spec and includes optional partner integrations.

Beautiful Rust is a package of three AI-agent skills (best-practices, review, port) plus a clippy lints config that teach coding agents to write, review, and port idiomatic Rust. It distills five published Rust style guides, works with major coding agents, and is validated by real model-run tests showing measurable improvement.

A set of Claude skills and local Python tools for Instagram content strategy, drafting, and optimization, featuring a humanizer to remove AI-generated patterns.

Moli is an open‑source, Rust‑based headless browser built for AI agents. It provides a full web runtime (HTML, CSS, JavaScript, networking, storage) but keeps memory and CPU low by performing layout and painting only on demand. The same binary can be used via a simple CLI, Chrome DevTools Protocol, or WebDriver Classic/BiDi, letting existing automation tools (Playwright, Selenium, etc.) connect without extra adapters. Benchmarks show Moli uses ~15 % of Chrome’s CPU time and ~13 % of its memory while achieving comparable success rates on real‑world web tasks. It is production‑ready, cross‑platform, and dual‑licensed (Apache‑2.0 or MIT).

Nuwa Skill is an open‑source Agent Skill that extracts the thinking patterns, mental models and expressive style of famous figures (Jobs, Musk, Naval, etc.) from public sources and packages them as portable “skills” usable in any AI‑agent runtime that supports the Agent Skills protocol.

A curated, community-maintained directory of plugins for the DeepSeek Harness agent framework, enabling users to easily discover and install extensions to customize their coding agent's capabilities.

A production-ready agent framework that provides an SDK and a FastAPI backend to build multi-tenant, multi-session agentic applications leveraging LLM reasoning and tool use.

A multi-agent tool that turns codebases and documentation into interactive knowledge graphs, helping developers visualize structure, dependencies, and business logic.

An open-source AI Development Environment (AIDE) that lets developers visually build, debug, and deploy production-ready AI pipelines using a high-performance C++ runtime.

A library of 22 packaged AI-agent workflows ("skills") for academic research — paper reading, Nature-style writing, figure generation, reviewer responses, literature search, and patent drafting — installable into Claude Code, Codex, Chatbox, and other agents.