Trending AI projects

Overlay Studio 是一个免费工作台，通过20种预调好的动效卡和AI语义编排，让口播视频的动效制作变成“挑卡对时间”而非“AI现场发挥”，并支持可视化编辑和透明层导出，直接导入剪映。

An open-source, 3D-printable robotic arm design (SO-101) designed for AI teleoperation and learning, fully compatible with the Hugging Face LeRobot library.

An open‑source Claude‑Code‑driven workspace that turns a natural‑language video brief into a full production pipeline—script, voice‑over, music, images, AI‑generated clips, transitions, and final MP4—using cheap open‑source models (Qwen‑3‑TTS, FLUX.2, ACE‑Step, LTX‑2) and cloud‑GPU services (Modal/RunPod). It provides slash‑command templates, brand profiles, a project lifecycle manager, and optional YouTube publishing, all under an MIT license.

A high-performance terminal workbench with persistent sessions and native integration for AI agents, allowing agents to be scripted and managed across multiple repositories.

VeRL‑Omni is an open‑source RL training framework for multimodal generative models (diffusion, unified, and omni‑modality). Built on verl and vLLM‑Omni, it provides fast rollout, asynchronous multi‑reward serving, distributed back‑ends, and many RL algorithms (FlowGRPO, DiffusionOPD, DPO, etc.). Supports GPUs and Ascend NPUs, with extensive docs, examples, and Apache‑2.0 licensing.

A browser extension that acts as a centralized management dashboard for AI API relay accounts, helping users track balances, automate check-ins, compare pricing, and sync credentials across various AI development tools.

NanoClaw is a lightweight, open‑source framework for running AI agents (Claude, OpenAI, Ollama, etc.) in isolated Docker containers. It provides secure per‑agent sandboxing, multi‑channel messaging (Slack, Discord, WhatsApp, etc.), a credential‑gateway for secret handling, scheduled tasks, and a code‑assistant‑driven setup/debug flow. The core codebase is intentionally small, encouraging users to customize behavior by editing code (or asking Claude Code to do it) rather than managing sprawling config files.

CLI Printing Press is a Go generator that builds token‑efficient CLIs and MCP servers for any API (even undocumented ones) by researching docs, sniffing traffic, and absorbing existing tools. It produces Go binaries with a local SQLite store, compound commands, and automated verification, aimed at AI agents like Claude Code.

QuantMind is an open‑source AI‑first quantitative‑trading platform that combines Microsoft Qlib, RD‑Agent LLM factor mining, and a chat‑bot (QuantBot) reachable from WeChat, QQ, Feishu, etc. It offers end‑to‑end data → factor → model → back‑test → simulated trading pipelines, 13 built‑in model types, autonomous factor evolution, and Docker‑based deployment.

QM is an open‑source, self‑hosted platform for multi‑user AI agents. It provides isolated per‑person workspaces and shared Slack/web channels, supports any LLM via plug‑ins, stores state in PostgreSQL, and offers fine‑grained security/posture controls. Deploy with the `qm` CLI, customize via a deployment repo, and extend with skills, sandboxed tools, and background jobs.

A command-line interface for Google Colab that allows users to provision cloud runtimes, execute code, and manage files directly from their terminal for automated workflows.

Codex‑autoresearch is a Git‑backed Codex skill that runs an autonomous loop: propose a single code change, commit it, measure a numeric metric (e.g., test failures), keep the change if it improves the metric and passes optional guards, otherwise revert. It repeats until a user‑specified target is reached, storing an immutable event log and HTML report. Supports foreground or background execution, works with any measurable outcome, and enforces strict safety (all trials are real commits, failures abort with clear errors).

Jonex is an enterprise-grade knowledge platform that converts raw multimodal data into structured, reasoning-ready knowledge graphs to power AI agents with accurate, traceable information.

The repository provides open‑source “AI Employees” – ready‑made AI agents that act as business staff (e.g., GTM Engineer, SEO, Web Dev, Social Media, etc.). Each employee is a folder of plain‑text schedules and routines that run on the user’s own machine via a compatible AI coding assistant (Claude Code or 12 other agents). Installation is a one‑line `npx ai-employees hire <role>` command or a guided Club wizard; the agent then automates daily tasks, drafts content, and delivers a concise morning brief, while all irreversible actions require manual approval. The project is MIT‑licensed, runs on the user’s existing Claude subscription, and includes extensive docs, guardrails, and a community club for support.

Apache Maka is a high-performance, local-first agent workspace that provides a transparent, auditable runtime for executing and benchmarking AI agents.

NVIDIA NemoClaw is an open‑source stack that runs AI agents (OpenClaw, Hermes, LangChain Deep Agents) inside NVIDIA OpenShell sandboxes, providing a CLI, managed inference, network‑policy enforcement, sandbox hardening, and state persistence for safer on‑premise agent execution.

A code-first Python framework for building, evaluating, and deploying sophisticated AI agents using graph-based workflows and modular multi-agent systems.

A real-time GPU video enhancement tool for Windows that applies super-resolution, denoising, and frame generation to local videos, capture cards, and game streams.

FastVideo is an open‑source framework for training, distilling, and running video diffusion models at real‑time speeds. It offers sparse‑attention back‑ends, LoRA finetuning, and a ready‑to‑use web app (Dreamverse) for streaming video generation, and runs on NVIDIA GPUs, Apple Silicon, and DGX Spark.

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.

ctf‑skills is a collection of Agent‑Skill modules that package CTF knowledge (web, binary, crypto, forensics, OSINT, etc.) and the required tooling into a format usable by LLM‑driven agents (e.g., Claude Code). It includes a one‑command installer for all common pentesting tools and an orchestrator (`/solve‑challenge`) that routes a challenge description to the appropriate skill, enabling AI agents to assist in solving CTF problems.

Xalgorix is an open‑source, self‑hosted AI‑driven penetration‑testing platform. An autonomous LLM agent runs a 22‑phase methodology, then a separate verifier re‑exploits each finding, delivering exploit‑verified results. Users bring their own LLM API key, run the tool via binary, Go install, Docker, or source build, and interact through a local web UI or API. Features include a bundled offensive‑security toolset, live telemetry, PDF reports, CI/CD GitHub‑App integration, and optional hosted SaaS. Licensed under Apache 2.0.

Cumula is a cross‑platform team chat where AI agents share the same roster, DMs, Kanban board and calendar as humans. Agents run either in managed Kubernetes pods (cloud) or on a user‑provided machine (BYOA) and can call tools, send real email, and coordinate without colliding. The stack uses a React/Vite UI, a stateless Express + WebSocket backend with Postgres and Redis, Cloudflare Workers for email/CDN, and Go FUSE for workspace mounting. Local development requires Postgres, Redis and an OpenAI API key; the repo includes full docs, benchmarks, and CI‑gated contributions.

Skyvern is an open‑source platform that adds LLM‑driven, vision‑based automation to Playwright. It lets developers and non‑technical users control browsers with natural‑language prompts, build reusable workflows, and run either locally (pip or Docker) or via a managed cloud service.

An agent-driven media library manager that automatically searches for movies, TV shows, and anime and saves them to your personal cloud storage.

llm‑for‑zotero is a Zotero add‑on that embeds a chat powered by any LLM (OpenAI‑compatible, Claude, Codex, etc.). It lets you ask questions about PDFs, get grounded citations, compare papers, save notes to Zotero or external Markdown folders, and (in beta Agent Mode) perform library‑wide actions like tagging, importing, and web search. Install via a .xpi file, configure a provider in Zotero’s preferences, and start chatting from the PDF reader.

MLX-VLM runs and fine-tunes vision-language and omni models (images, video, audio) locally on Apple Silicon using the MLX framework, with CLI, server, chat UI, speculative decoding, and conversion tools.

An open-source, self-hosted omnichannel marketing and AI agent platform that allows businesses to build automated chat flows and deploy LLM-powered agents across multiple messaging services.

A Claude Code Skill that converts content from 15+ sources, including paywalled news and podcasts, into AI-generated podcasts, PPTs, and mind maps via Google NotebookLM.

LibreTranslate is a self-hosted, open-source machine translation API that removes dependency on proprietary services like Google or Azure.

Better DeepSeek is an open‑source browser extension that enriches the DeepSeek chat web app with custom system prompts, file‑generation tags (PowerPoint, Excel, Word, code, visualizers), persistent local memory, voice interaction, and experimental code‑planning (DeepCode) that can hand off tasks to DeepSeek Harness. It runs entirely in the browser, is installable from Chrome/Edge/Firefox stores or via manual build, and is aimed at power users who want richer, private AI‑assistant capabilities.

A self-hosted tool runtime that gives cloud AI agents secure access to a local development environment, allowing them to edit code and run local tools without moving the repository.

An MCP server and CLI tool for searching and downloading academic papers from multiple open-access and public sources using a unified interface.

Polars is a high-performance analytical query engine for DataFrames that uses a multi-threaded, lazy execution model to process large datasets efficiently, even those exceeding available RAM.

A self-hosted CLI that uses an LLM to turn a product description into searched, qualified B2B leads (each with a written reason), resolves their work emails, and drafts and sends the outreach email from your own mailbox — with zero scraping or social-platform accounts.

FastMCP is a framework that simplifies building Model Context Protocol (MCP) servers and clients, allowing developers to easily connect LLMs to external tools and data using simple Python code.

CodeAF is an open‑source Go tool that turns LLMs into a multi‑project coding factory. From a single terminal window you can chat, have the system plan, run and check code across many repositories, using a crew of three model roles per task. It auto‑selects the cheapest open model, supports specialist sub‑harnesses, offers a headless CLI for CI, and runs with a tiny binary footprint.

A local-first, source-available AI meeting assistant that provides real-time transcription, context-aware answers, and advanced stealth features for interviews and professional calls.

Kimodo is a motion diffusion model that generates high-quality 3D animations for humans and robots, controlled via text prompts and precise kinematic constraints.

A command-line state machine that turns AI agent workflows into inspectable graphs of states and executable checks to ensure reliability in long-running tasks.

Occamy-1.0 is an open-source, compact agentic language model (35B total, 3B active) designed for long-horizon co-work tasks like tool use, file manipulation, and multi-step workflows. It achieves leading results on co-work benchmarks while keeping inference efficient.

A browser automation skill for AI agents that provides web connectivity, CDP-based browser control, and the ability to use existing browser sessions and local history.

Gravitas is an open-source reliability runtime for AI coding agents that enforces engineering discipline by requiring agents to ground their edits in evidence and verify their work through automated tests before finishing a task.

OpenEnv is an end-to-end framework for creating, deploying, and using isolated execution environments for agentic reinforcement learning, with a simple Gymnasium-style API. It provides a unified interface to diverse environments (games, coding sandboxes, browsers) as Docker images, enabling training with major RL frameworks.

A unified CLI and Python API that provides an easy-to-install wrapper for multiple Whisper AI transcription backends, featuring isolated environments and speaker diarization.

MAPIE is a Python library that provides model‑agnostic prediction intervals and risk‑controlled prediction sets for regression, classification, time‑series, and computer‑vision tasks. It works with any scikit‑learn estimator (and can wrap TensorFlow/PyTorch models) and offers statistically guaranteed uncertainty estimates based on conformal prediction. Install via `pip install mapie` and use the `MapieRegressor` or `MapieClassifier` wrappers to obtain calibrated intervals/sets with minimal code.

Prismor is a runtime security and governance platform for AI agents that monitors and enforces policies on agent actions, tool calls, and network access to prevent security risks like prompt injection and data exfiltration.

A curated collection of “Agent Skills” that teach how to manage an LLM’s limited context window. The repo organises the material into self‑contained skill folders (foundational, architectural, operational, etc.) that can be loaded by Claude Code, Cursor, or any Open‑Plugins host. It explains context‑engineering concepts (lost‑in‑the‑middle, compression, self‑editable context), provides platform‑agnostic examples, and includes benchmark results showing high skill‑router accuracy on frontier models.