Trending AI projects

Agents' Last Exam (ALE) is an open‑source benchmark framework that evaluates AI agents on 55 real‑world professional tasks across 13 industry clusters. It provides sandboxed Linux/Windows environments (cloud VMs, Docker, QEMU/KVM, etc.), a run orchestrator (`ale_run`) that hands a task description to the agent, lets it work autonomously, then grades the output against a hidden reference. Scores are reproducible, and logs capture every tool call and screenshot. The project is led by UC Berkeley RDI, released under Apache‑2.0 (code) and CC‑BY‑4.0 (data), with a website, leaderboard, and community resources.

nostr-vpn is a Tailscale-style private mesh VPN using Nostr identities for enrollment and a FIPS-based data plane, with an experimental Bitcoin-paid exit-node marketplace. It offers a CLI, daemon, and native apps across desktop and mobile, plus server packages for StartOS and Umbrel. The fork claims ~3× throughput over the original FIPS implementation. Enrollment requires explicit admin approval; paid exits use Cashu channels with per-byte pricing.

An AI-native, one-stop writing tool for fiction authors that uses Agentic RAG and customizable workflows to help maintain complex world-building and character consistency.

A voice interface for Claude Code and other MCP-capable agents that enables natural, hands-free conversations using either cloud or local speech services.

A curated, timeline-ordered list of open-source AI web search tools that combine large language models with web capabilities, covering summarization, function calling, and agent-driven research.

Gloss is an open‑source web app that lets you ask questions about any selected line in an article or PDF and get AI‑generated, cited answers right in the margin, using Together.ai LLMs, Exa search, and client‑side PDF parsing.

A searchable, daily‑updated catalog of Claude Code plugins with web UI, CLI (`sk`), and JSON API; includes security scanning, deduplication, and a canonical taxonomy.

The repository provides a collection of 52 AI‑driven “skills” that turn any LLM into a full‑stack affiliate‑marketing assistant, covering research, content creation, landing‑page building, analytics and automation, with a closed‑loop flywheel and a tiny CLI for testing.

tunnel-client is an OpenAI‑provided daemon that creates a secure, OpenAI‑hosted tunnel so a private MCP (Model Context Protocol) server can be reached by ChatGPT, Codex, and other OpenAI services without exposing the server to the public internet. It offers health/metrics endpoints, role‑based permissions, embedded demo modes, and a Go SDK for in‑process use, and can be installed via Homebrew, Docker, or Kubernetes.

A CLI text-to-speech tool using the Kokoro model that converts text, PDFs, and EPUBs into natural-sounding speech with support for voice blending and streaming.

A set of API documentation and SDKs that allow the AI personality Neuro-sama to play video games by communicating game states and actions via a websocket protocol.

OpenMVS is a computer vision library that completes the photogrammetry pipeline by transforming sparse 3D point clouds and camera data into detailed, textured 3D meshes.

A security-focused orchestration engine that automates complex security workflows and reconnaissance using declarative YAML definitions and distributed execution.

A reverse-engineering project that detects and removes Google's invisible SynthID watermarks from Gemini-generated images using spectral analysis and a multi-stage adversarial pipeline.

Browser Control is an open‑source Node.js driver that lets AI agents or CLI users run Playwright code inside their own Chromium browser (with all logged‑in sessions and extensions). It works via a local relay and a browser extension, supports persistent sessions, inspection helpers, human hand‑offs, video/ HAR recording, and a TypeScript Effect client for secret‑safe API calls. Intended for trusted local use only.

Mamba is a linear-time sequence modeling architecture based on selective state space models (SSMs) that provides a high-performance, subquadratic alternative to Transformers.

A PyTorch-based open-source framework for deep learning in healthcare imaging that provides standardized workflows and domain-specific tools for medical data.

pxpipe is a local proxy that compresses bulky LLM request parts (system prompts, tool docs, long histories) into PNG images, dramatically reducing token usage for vision‑enabled models like Claude Code and Gemini. It works by rendering text blocks to images, sending those images to the model, and preserving only a limited set of exact identifiers as text. The tool includes a dashboard, a warp wrapper, offline export utilities, and a statistics reporter. Savings of 59‑70 % of the total bill have been reported on dense workloads, though the conversion is lossy for exact hex strings and works best when the request is token‑dense.

A real-time object detection framework based on the Detection Transformer (DETR) architecture that aims to outperform YOLO models in speed and accuracy.

A community-maintained Linux distribution of the official ChatGPT desktop app that adds native package support and a wide range of optional system-integration features.

CubeCL is a Rust language extension and just-in-time compiler that lets you write compute kernels once in Rust and run them on CUDA, HIP, Metal, Vulkan, WebGPU, or CPU. It combines automatic vectorization, comptime specialization, and autotuning to deliver high performance across very different hardware backends.

MiniMind‑O is a genuine open‑source, tiny (≈0.1 B) end‑to‑end omni model that jointly handles text, speech and images as inputs and produces text and streamed speech outputs. It provides full PyTorch code, training scripts, mini/full datasets, pretrained weights, voice‑cloning prompts, and demo interfaces (CLI, Gradio, phone‑mode UI). The repository is clearly an AI/ML project, not an off‑topic list.

BoxLite is a secure, daemonless compute engine that provides hardware-isolated micro-VMs for AI agents to safely execute code and persist state across tasks.

A cross-platform AI chat client that unifies local on-device LLM inference via llama.cpp with cloud API access for a flexible, privacy-focused chat experience.

An efficient LLM inference framework optimized for Chinese AI accelerators to enable high-throughput, low-latency enterprise deployment.

An AI-native database providing high-speed hybrid search across dense and sparse embeddings, tensors, and full-text for LLM and RAG applications.

A lightweight 2D digital human engine that enables real-time animation and dialogue on mobile browsers and low-power devices without requiring a GPU.

An MCP server that enables AI models to search DuckDuckGo and fetch webpage content with built-in bot detection bypass and context-saving URL shortening.

AIMET is Qualcomm’s open‑source toolkit for quantizing and compressing PyTorch/ONNX models, enabling fast, low‑memory inference on edge devices with minimal accuracy loss.

A Python package that integrates AI frameworks with geospatial data analysis, providing tools for satellite imagery processing, model training, and inference.

Agent Chat UI is a Next.js front‑end that lets you chat with any LangGraph server. Configure the server URL, assistant ID, and optional LangSmith key, then use the web UI to send messages, stream LLM replies, hide unwanted output via special tags, and display extra “artifacts” in a side panel. The repo includes a quick‑start (`npx create-agent-chat-app`), environment‑variable shortcuts, and production guidance (API‑passthrough proxy or custom authentication).

A high-performance, cross-platform C++ library for image and video beauty filters using OpenGL/ES.

Flowboard is a local‑only, graph‑based UI for building e‑commerce media pipelines that generate images and videos via Google Flow (Veo 3.1 i2v). It uses a Chrome extension to proxy signed Flow requests and a locally installed LLM CLI (Claude, Gemini, or Codex) for vision and auto‑prompting. All board data is stored in SQLite; no cloud backend is required.

JimuReport is a free, open‑source Java‑based BI suite (Spring Boot starters) that lets users design reports, dashboards, and big‑screen displays via drag‑and‑drop. It adds an AI assistant (Claude Code or any OpenAI‑compatible model) that can generate or modify visualisations from natural‑language prompts, including a conversational Chat2BI interface. Supports 30+ data sources (including Chinese “信创” databases), mobile rendering, printing, and data entry. Licensed under LGPL 3 with optional commercial licensing.

Mobile Harness is a Python‑based, Markdown‑driven library that lets LLM agents control Android and iOS devices (locally via ADB/Portal or remotely via Mobilerun cloud). It offers a unified API for UI interaction, screenshots, app launch, and accessibility‑tree queries, and includes benchmarks showing near‑perfect performance on AndroidWorld and iOSWorld task suites.

An infrastructure layer that provides a single, shared persistent memory database with vector search for all your AI tools, eliminating fragmented data across different AI clients.

Resume Optimizer is a skill for AI coding agents that audits and rewrites resumes to be achievement-focused and tailored to a target job description, turning duty descriptions into results, surfacing metrics, flagging common risks, and optionally compressing to one page.

A collection of high-performance GPU kernels for deep learning operations like RMSNorm and Softmax, implemented using CuTe-DSL for modern NVIDIA GPUs.

A collection of ComfyUI nodes that enables users to import, manipulate, and export video files and image sequences, providing essential tools for frame-rate adjustment, batch processing, and video encoding.

A comprehensive Chinese learning resource for DeepLearning.AI's Agentic AI course, providing translated content, structured knowledge breakdowns, and detailed code analysis to help users build autonomous AI agents.

A high-performance inference toolkit for deploying and serving open-source text embeddings and sequence classification models with low latency.

A collection of structured, open-source Agent Skills compatible with the Agent Skills standard, providing reusable workflows for content creation, image generation, and AI project engineering.

A TypeScript and JavaScript SDK that makes functions durable by persisting progress, retrying failed steps, and allowing workflows to suspend without using compute.

Pi 学习蓝皮书是一本面向中文初学者的非官方学习手册，提供从安装、登录到第一个可验收任务的完整路线，帮助用户逐步掌握 Pi Coding Agent 的会话、上下文、扩展和长期工作流，并强调独立验收而非轻信 Agent 自述。

An agent skill that enables AI agents to produce institution-grade research reports and white papers by applying conventions reverse-engineered from top-tier global organizations.

A curated, community-maintained collection of research papers and resources focused on Embodied AI, including Vision-Language-Action models, world models, and robotic navigation.

Awesome Quant AI is a curated resource list (papers, tools, books) for AI-driven quantitative trading, plus original explainer content covering strategy design, strategy types, paradigm comparisons, and emerging 2025/2026 topics like LLM trading agents and time-series foundation models.

A curated npm package that bundles hand-picked agent skills from across the ecosystem, so projects get them installed and auto-synced with a single dependency.