Trending AI projects

A portable agent skill that provides AI coding tools with expert SwiftUI best practices and a toolchain for analyzing performance traces.

An open-source, self-hosted construction ERP platform that integrates BIM, cost estimation, and project management with AI-assisted workflows.

NextChat is an open‑source, web‑based AI chat UI that works with many LLM providers, offers privacy‑first local storage, prompt templates, plugins, and cross‑platform desktop clients. Deployable in one click on Vercel or via Docker, it also provides an Enterprise edition with branding and admin controls.

1Cat‑LLM is a vLLM fork that optimises LLM inference for NVIDIA Volta (Tesla V100) GPUs, offering custom kernels, FP8/E4M3 KV‑cache, GGUF support and speculative decoding to achieve up to ~260 tokens / s on 27‑B models while reporting quality‑aware benchmarks.

An open-source AI operating system that coordinates a lead agent and Docker-isolated worker agents to automate complex workflows across Slack, GitHub, and email.

vLLM Ascend is a hardware plugin that enables the vLLM LLM serving engine to run on Huawei Ascend NPUs. It supports a range of Atlas NPU hardware, requires Linux, Python 3.10‑3.12, CANN 9.1.0, PyTorch 2.10.0 + TorchNPU, and matches specific vLLM versions. The project provides regular releases, detailed docs, and an active community (Slack, forum, weekly meetings). It’s licensed under Apache 2.0.

Gymnasium is a community‑maintained Python library that provides a standard, versioned API for reinforcement‑learning environments (classic control, Box2D, Atari, MuJoCo, etc.). It’s a fork of OpenAI’s Gym and serves as the main platform for developing and benchmarking RL algorithms.

Genie Sim is a comprehensive, agent-friendly simulation platform for embodied AI that provides tools for scene reconstruction, large-scale synthetic data generation, and standardized robot performance evaluation.

A curated, tightly scoped directory of 2,162 production-ready APIs for building autonomous AI agents, organized into Agents, AI Models, and MCP Servers. It helps developers quickly shortlist the right APIs for agentic systems instead of wading through unrelated categories.

An AI development framework that transforms natural language ideas into complete software projects by automating requirements gathering, task planning, and multi-agent execution.

PiG is a Go‑native reimplementation of the Pi terminal coding‑assistant, offering a fast single‑binary AI pair‑programmer compatible with Pi’s extensions and configurable via “Piglets.”

RedAmon is an autonomous, Docker‑based security platform that chains parallel reconnaissance tools, feeds their results into a Neo4j knowledge graph, and uses LLM agents to triage findings, validate exploits, and automatically generate GitHub pull‑requests with code fixes. It supports many OSINT APIs, multiple LLM providers, optional OpenVAS integration, and provides a web UI for admin‑controlled, human‑in‑the‑loop operation.

A cross-platform desktop management tool for OpenAI Codex that provides a visual interface for managing prompt templates, API providers, sessions, and Skills/MCP extensions.

Pentest Swarm AI is an open‑source, self‑hosted framework that runs dozens of independent AI agents in parallel to discover, exploit, and report vulnerabilities. It uses a stigmergic blackboard with pheromone‑based scheduling, works with any LLM (Claude, Together AI, Ollama, etc.), includes a built‑in recon toolchain, live web/TUI dashboards, and generates evidence‑backed reports. Install via npm, Homebrew, Go, script, or Docker, then launch with `pentestswarm run`. Licensed under AGPL‑3.0 and currently in alpha.

Operit2 is a cross-device agent framework that synchronizes tasks, context, and tool execution across personal devices like phones, desktops, and cloud servers.

A 3D virtual assistant SDK for Unity that turns 3D VRM models into voice-enabled chatbots with synchronized speech, lip-sync, and autonomous animations.

gemini‑web2api is a Python server that proxies Google Gemini’s private web API and presents it as an OpenAI‑compatible endpoint (`/v1/chat/completions`, `/v1/models`, etc.). It supports optional API‑key auth, streaming, function calling, image input, multiple Gemini model aliases, adjustable reasoning depth, and can run directly or inside Docker.

A library of production-ready AI workflows and automation scripts designed to be used with AI coding agents to handle marketing, sales, and content operations.

ZenConverter is a private, local-first file conversion app for Android that processes media, documents, and images directly on your device without requiring cloud uploads or internet access.

An independent, community-curated directory of skills, plugins, tools, and resources for the Hermes Agent, with maturity tags to help users pick reliable add-ons and a trust-boundary guide for safe adoption.

An open-source desktop app that uses AI agents and a local SQL engine to analyze and actually operate on social chat histories from multiple platforms privately.

A self-learning system for Claude Code that captures user corrections and discovers workflow patterns to create permanent memory and reusable skills.

A complete offline AI suite for Android, iOS, and macOS that provides text, image, and vision AI capabilities entirely on-device for total privacy.

safelabs‑eval is an open‑source Python framework that red‑teams AI agents. It sends 300 curated adversarial prompts (30 per OWASP‑inspired ASI01‑ASI10 categories) to a target agent—via HTTP or a Python callable—and scores the replies with pure‑Python pattern detectors. The tool is framework‑agnostic (supports LangChain, CrewAI, AutoGen, LlamaIndex, OpenAI Agents SDK, Google ADK, Semantic Kernel via optional adapters), produces human‑readable or JSON reports, and requires no extra LLM calls. It helps developers discover prompt‑injection, data leakage, tool‑use safety, hallucination, and other agent‑specific risks before deployment.

A personal business operating system for solopreneurs that uses AI agents to manage various business departments from a single unified command center.

Talorys is an open‑source personal AI assistant that runs entirely inside your own Cloudflare account. It offers chat, memory, task/note/project management, and scheduled automations, storing everything in a SQLite‑backed Durable Object with no external servers.

A collection of distilled software engineering rules from classic programming books designed as project-level instructions to guide AI coding agents in producing high-quality architecture and code.

DBHub is a minimal, token‑efficient MCP server that lets LLM‑based tools query PostgreSQL, MySQL, SQL Server, Oracle, SQLite, and MariaDB safely. It provides only two core tools (SQL execution and schema search) in ~1.4 k tokens, optional extra tools, guardrails, and a built‑in web UI, making it a lightweight bridge between AI agents and real databases.

Kairos is a 4B-parameter native cross-embodiment world model that unifies visual generation and executable action prediction for robots across different hardware platforms.

Claude‑Skills‑Founder is a Claude Code plug‑in that provides founder‑focused AI commands (idea validation, competitor matrix, pricing, pitch‑deck, etc.). It stores each output in a `founder/` folder so later commands can reuse the data, enabling a structured, repeatable early‑stage workflow.

Megatron-LM is a GPU-optimized library and reference framework designed to scale the training of massive transformer models across large clusters using advanced parallelism and precision techniques.

A unified framework to evaluate generative language models on a large number of standard academic benchmarks, serving as the backend for the Open LLM Leaderboard.

Agents Office is a beta, locally‑run visual platform that turns Claude into a team of 35 role‑based AI agents. You type tasks, the system routes them to the right agent (or a team), the agents use Claude Code connectors (Gmail, Slack, Chrome, etc.) to do the work, and results are saved as notes in your personal “Brain”. Features include customizable briefs/skills, recurring routines, a calendar view, team‑splitting, model/effort selection, and a non‑commercial license.

A collection of high-quality SKILL.md files for AI agents (Cursor, Claude Code, etc.) that enable automatic task triggering for academic writing, diagramming, and development workflows.

A fast BPE tokenizer for OpenAI models that converts text into tokens for model processing with high efficiency and lossless compression.

A tool for automatically generating themed image cards from Markdown and publishing them as notes to Xiaohongshu.

A terminal application that manages multiple AI coding agents in isolated git workspaces, allowing developers to run several tasks simultaneously without conflicts.

QVAC is a local-first AI SDK and model provider that enables developers to run LLMs, vision, audio, and generative models on any device while supporting peer-to-peer model distribution.

An AI-native API gateway based on Istio and Envoy that provides unified LLM API management, MCP server hosting, and production-grade traffic control for AI workloads.

Lody is an open‑source, cross‑platform workspace that lets teams share and coordinate AI coding agents (Claude Code, Codex, Kimi, etc.) across desktop, mobile, web and CLI. It syncs conversations, Git worktrees, files and visual feedback using CRDTs, enabling collaborative, version‑controlled AI‑driven development.

ACRYL is an early-stage (v0.2.5, MIT-licensed) "Agentic Development Environment" that keeps your project, context, and work persistent while letting any coding agent (Claude Code, Codex, etc.) plug in as a replaceable worker. It features a canvas-based UI, a catalog of 33 agents, agent-driven app control with per-action approval, a plugin system built on Cordis, shareable "Blend" environments, and three interchangeable surfaces (desktop, local web, terminal). It builds on a pinned, unmodified DeepSeek Harness checkout as its capability substrate.

Nightingale is an open‑source, self‑hosted alerting platform that aggregates alerts from multiple monitoring data sources (Prometheus, VictoriaMetrics, Elasticsearch, Loki, ClickHouse, PostgreSQL, etc.). It provides a UI/API for rule management, team‑based permissions, event processing pipelines, and native notification integrations (Slack, PagerDuty, Discord, Email, Telegram, Jira, …). A built‑in AI assistant can query monitoring data, explain alerts, generate or modify rules, and run remediation scripts via extensible “Skills”. Deployable via Docker Compose, Linux binaries, or Helm on Kubernetes, Nightingale is Apache‑2.0 licensed and backed by a community Slack and GitHub discussions.

nndl/nndl 是《神经网络与深度学习》一书的公开仓库，提供第二版（系统理论型）和通识版（直觉型）两本电子书的免费 PDF、目录、勘误入口，以及配套的 PyTorch 实践代码和《大模型与智能体》专题读本链接。仓库本身主要存放主站元数据，不含第二版 LaTeX 源码。

kagent is a CNCF‑incubated, open‑source framework that lets you define and run AI agents as native Kubernetes custom resources. Agents are described in YAML, can use any major LLM provider, run on Go/Python runtimes or custom back‑ends, and have built‑in tool integration, persistent sessions, and OpenTelemetry tracing. The system includes a controller, web UI, CLI, PostgreSQL storage, and a substrate that manages agent lifecycles, making AI workloads manageable with standard Kubernetes/GitOps workflows.

deja-vu is a local, private memory engine for coding AI agents. It indexes transcripts from 41+ agent harnesses (Claude Code, Codex, Cursor, etc.), stripping secrets, and provides instant search, recall, and MCP integration so agents remember past fixes, commands, and file changes across sessions and tools. It features a fast CLI/TUI, proactive session-start recall, and benchmarks showing 97.2% R@5 on LongMemEval-S.

The README describes an open-source, community-curated directory of links for students and early-career developers — covering courses, competitions, scholarships, AI/ML, blockchain, security, and more — with emoji markers for difficulty level and cost, and a quick-start guide to help readers navigate it quickly.

OpenHarness is an open‑source native desktop hub that runs multiple LLM coding agents (Claude Code, Codex, etc.) in real tmux terminals, links laptops, servers and GPU boxes via end‑to‑end‑encrypted WebRTC/relay, and lets you add domain‑specific “harnesses” (Blender, CircuitJS, MuJoCo, etc.) with live viewers. A optional ESP32‑S3 round device shows agent status and accepts voice/tap input.

A lightweight, GPU-accelerated framework for robot learning that combines Isaac Lab's manager-based API with MuJoCo Warp.