Tracecat is an open‑source, AI‑native security automation platform. It lets teams build custom LLM agents with tool access, manage incident cases, and orchestrate deterministic workflows (via Temporal). Features include sandboxed code execution, 100+ security‑tool integrations, a Model‑Code‑Prompt (MCP) server for turning prompts into actions, and a self‑hostable stack (Docker, Fargate, Kubernetes). The core is AGPL‑3.0; an Enterprise edition adds multi‑tenant, RBAC, and human‑in‑the‑loop approvals.

An open-source signal processing machine learning toolkit based on PyTorch that simplifies the creation, augmentation, and transformation of RF signals for ML research.

pyscn is a fast, Go‑based static analysis tool for Python that scores code health, finds dead/duplicate code, measures complexity, and checks architecture. It integrates with AI coding agents via installable Skills and an optional MCP server, letting assistants run quality checks and suggest refactors directly from the IDE or chat.

OpenClaudia is a free, MIT‑licensed collection of 77 markdown‑based “skills” that let an AI coding agent (Claude Code, Codex, etc.) perform end‑to‑end marketing tasks—SEO audits, blog writing, ad copy, email sequences, social‑media posting, analytics, and more—by invoking simple slash commands. Install with `npx openclaudia install --all`, configure any needed API keys, and the agent can generate, publish, and send marketing assets locally, offering a fully customizable alternative to paid AI‑marketing SaaS.

Aiden is an open‑source autonomous AI engine (Node/TypeScript) that turns a natural‑language goal into a verifiable, multi‑step workflow across files, terminal, browser, APIs, and Git. It records every action in SQLite, supports risk‑tiered approvals, multiple LLM providers (OpenAI, Anthropic, Gemini, Ollama, etc.), and can run as a CLI, web workbench, or background daemon. Install via `npm i -g aiden-runtime` and start with `aiden`.

Everywhere is a cross‑platform desktop AI assistant (Windows/macOS) built with .NET 10 and Avalonia. It captures on‑screen context via accessibility APIs and lets you query LLMs (OpenAI, Anthropic, Gemini, etc.) through a global hot‑key, returning markdown‑rich answers. Features include browser/file/terminal agents, rich UI, multi‑language UI, and support for local Ollama models. The project is open source under its LICENSE and provides installers, a portable zip, and full build instructions.

ChatWiki is an open‑source, Docker‑based platform that turns a WeChat official account into an AI assistant. It offers visual workflow building, automatic replies (text, voice, images, mini‑program cards), knowledge‑base creation from scraped articles or uploaded docs, hybrid vector‑graph search, and human‑hand‑off. The backend uses Go and Python, the frontend Vue.js, and PostgreSQL 16 with pgvector for embeddings. It supports over 20 LLM providers (e.g., OpenAI, Claude, DeepSeek) and can be deployed as a web app, desktop client, or embedded in other services. The project is licensed for free personal use; commercial users need a paid license.

RHVoice is an open-source speech synthesizer that uses statistical parametric synthesis to provide lightweight, intelligible text-to-speech across multiple languages and platforms.

AnyCrawl is an open‑source, Node.js/TypeScript toolkit for fast, scalable web and SERP crawling. It offers multi‑process workers, three scraping engines (cheerio, Playwright, Puppeteer), batch jobs, Redis‑backed queues, caching, proxy support, and optional LLM‑driven JSON extraction. You can self‑host via Docker or use the public API, generate API keys, and call endpoints for single‑page scrape, full‑site crawl, SERP search, or batch scraping. Ideal for AI agents, data‑collection pipelines, and SEO automation. MIT licensed.

Sakura Desktop Pet is an open‑source AI‑driven desktop companion that watches the screen, remembers interactions, and can start conversations on its own. Users load a character pack (portrait, role‑card, optional voice) and configure LLM APIs; the agent then uses screenshots, tool calls (web search, reminders, notes) and a long‑term memory to act like a proactive virtual pet.

An end-to-end speech processing toolkit built on PyTorch that provides unified, reproducible recipes for ASR, TTS, speech translation, and enhancement.

A speaker diarization pipeline that combines OpenAI Whisper for transcription with NVIDIA NeMo for speaker identification to determine who spoke which words in an audio file.

jt‑live‑whisper is a cross‑platform desktop tool that captures system audio and runs Whisper‑based speech‑to‑text, local translation, speaker diarisation, and LLM‑powered meeting summarisation entirely on the user’s machine. It works on macOS, Windows and Linux, supports real‑time subtitles (including bidirectional English‑Chinese or Japanese‑Chinese), batch processing of audio files, floating overlay subtitles, keyword alerts and can forward subtitles to chat apps. Installation is a one‑click script that sets up Python, required libraries and the AI models; the app can be started via an interactive menu, a lightweight web UI, or direct CLI flags.

A general-purpose speech recognition toolkit that wraps Whisper models into a REST API for easy deployment and integration.

An open-source speech-to-text inference framework using PyTorch for deploying pre-trained transducer and CTC-based models.

A speech-practice app for iOS and Android that provides live feedback and AI coaching to help users improve their articulation, pacing, and fluency.

Warden is an open‑source Sentry project that runs AI‑powered code reviews using reusable “skills”. It can be invoked locally via a CLI or automatically on each PR through a GitHub Action, posting findings as inline comments and Checks. Supports OpenAI and Anthropic models and can auto‑apply simple fixes.

TornadoVM is an open‑source Java runtime that JIT‑compiles Java bytecode into GPU kernels (CUDA, OpenCL, Metal) and integrates NVIDIA libraries (cuBLAS, cuFFT, cuDNN) and Tensor‑Core intrinsics. It lets you write kernels in pure Java, build task graphs, and run them on NVIDIA, AMD, Intel, Apple‑Silicon GPUs or CPUs. Used for LLM inference (GPULlama3.java), ray‑tracing, and scientific pipelines, it is installable via SDKMAN! and available on Maven Central under Apache 2.0 (API) and GPLv2 + Classpath Exception (runtime).

An offline voice recognition input method for Android based on the Whisper engine, providing privacy-focused speech-to-text and translation.

An open-source voice AI orchestration platform for agencies and enterprises to build and deploy scalable, real-time voice agents without vendor lock-in.

Optio is an open‑source, self‑hosted platform that orchestrates AI agents (Claude Code, OpenAI Codex, Gemini, etc.) as durable sessions on Kubernetes or local machines. A single form captures *when, where, who, what, then* to create PR‑automation pipelines, scheduled jobs, interactive terminals, or long‑living agent swarms. It provides a unified UI/feed, Kubernetes‑style reconciliation, multi‑vendor model support, fine‑grained connections (Slack, Notion, GitHub, etc.), cost tracking, and enterprise‑grade security—all running inside your own infrastructure.

AgentGuard is an npm‑distributed security layer for AI coding agents (Claude Code, OpenClaw, Hermes, etc.). It installs once, hooks into the agent runtime, and blocks destructive commands, protects secret files, and logs which skill triggered each action. It also offers static‑analysis scans for new skills, a daily “patrol” health‑check (8 built‑in checks), and optional cloud‑connected threat‑intel. Installation is `npm i -g @goplus/agentguard && agentguard init`; usage includes `agentguard scan`, `agentguard protect`, `agentguard patrol`, and `agentguard checkup`.

noteDigger is a pure‑JavaScript web app for audio‑to‑MIDI transcription. It lets users upload audio/video, visualises a spectrogram (STFT/CQT, GPU‑accelerated), runs on‑browser neural‑network models for pitch and instrument‑separation transcription, and provides extensive manual editing tools with keyboard shortcuts. Users can export quantised MIDI files or save a project for later work, all without external libraries or server‑side processing.

AgentPet is a cross‑platform desktop app that monitors the live state of AI coding agents and gamifies the experience with a tamagotchi‑style pet that levels up based on token usage and completed sessions. It works offline, supports many agents, offers per‑project dashboards, and provides optional web sync and leaderboards.

LLM Sandbox is a Python package that runs code generated by large language models inside isolated containers (Docker, Kubernetes, or Podman). It supports multiple languages, enforces security policies, can capture output plots, offers interactive notebook‑style sessions, and includes a container‑pooling system for fast, concurrent execution.

Moss is an open‑source Python framework that turns plain‑language crypto‑trading ideas into back‑tested, self‑evolving quantitative agents. It parses natural‑language descriptions, builds a strategy with technical indicators, runs local Hyperliquid back‑tests, and automatically refines parameters each week while enforcing safety guardrails. Optional integration lets you upload results to the Moss platform and run live bots on Hyperliquid. The project is MIT‑0 licensed and intended for research/educational use.

A collection of generative models from Sony AI for creating high-quality sound effects via text-to-audio and video-to-audio generation.

A real-time interpretation platform for live events that allows interpreters to broadcast translated audio to attendees via a browser-based, low-latency stream.

K8E is an open‑source, single‑binary sandbox platform that lets AI agents run untrusted code securely on any Linux host. It supports gVisor, Kata, or Firecracker isolation, provides per‑session eBPF network policies, warm‑pool pods for sub‑500 ms start‑up, and an E2B‑compatible API accessed via the `k8e-sandbox-cli` tool.

ComfyUI nodes for OmniVoice, providing zero-shot multilingual text-to-speech with voice cloning, voice design, and multi-speaker dialogue support for over 600 languages.

A private, on-device voice dictation and transcription tool for macOS that uses local Apple frameworks and LLMs to provide speech-to-text, translation, and text refinement.

API Relay Audit is a zero‑dependency Python script that lets you locally test any third‑party AI‑API relay or LLM proxy for hidden prompt injection, context truncation, model substitution, tool‑call rewriting, and Web3‑wallet safety. It runs a 14‑step audit, outputs a Markdown report with LOW/MEDIUM/HIGH risk ratings, and can be used as a standalone script or as a DeepSeek Harness plugin.

exxperts is a cross‑platform desktop/CLI app that runs AI agents locally with opt‑in, auditable memory. Rooms act as persistent workspaces; after each chat the agent proposes facts to remember, which you must approve. All data stays on your machine, supports any OpenAI‑compatible model, includes built‑in DuckDuckGo search, and is released under Apache 2.0.

nWave is a Claude Code plugin suite that orchestrates AI agents through a seven‑step “wave” workflow (discover → deliver). Each wave produces reviewed artefacts, and a DES layer enforces TDD, peer review, and configurable rigor profiles. Install via a one‑line script (Python 3.10+, uv or pipx), then drive the process with slash commands like `/nw‑design` or `/nw‑buddy`. Designed for developers and teams that want AI‑assisted coding while keeping human oversight at every gate.

A research project from Microsoft Research Asia that provides a suite of deep learning models for music understanding, generation, and AI-driven music processing.

A Python text-to-speech library designed for low-latency audio generation from strings, generators, and LLM token streams, supporting a wide variety of local and cloud TTS engines.

A lightweight and highly accurate wake word engine for building always-listening voice-enabled applications across IoT, mobile, and web platforms.

A Windows desktop app that turns system audio into live subtitles and translations using various speech-to-text engines and local offline models.

A toolkit for training and running inference on audio generation models, supporting various model types including diffusion and autoencoders.

claude‑replay is a zero‑dependency tool that converts transcript logs from Claude Code, Cursor, Codex CLI, Gemini CLI, OpenCode, Kimi Code, and Hermes Agent into a single, self‑contained HTML replay. The replay is interactive (play/pause, speed control, toggle thinking/tool blocks), themeable, and can be edited via a built‑in web editor. It supports live watch mode, secret redaction, and easy embedding, making AI‑coding sessions shareable for blogs, docs, demos, bug reports, and teaching.

A Python MCP server that exposes Yahoo Finance data (prices, statements, options, news, holders, analyst ratings) as typed tools for LLMs like Claude, enabling AI‑driven stock and market research.

HarnessFlow is a portable set of Markdown instruction files that turn Claude Code, Codex CLI, or GitHub Copilot into a multi‑step, self‑reviewing coding assistant. It provides request templates, parallel analysis agents, a devil’s‑advocate challenge, QA validation, and persistent repo memory. Install by copying the pack into `.github/HarnessFlow/` of your repo and running the provided `setup.sh` or `cli_setup.sh`. Choose a request template (code, refactor, debug, etc.), fill it in, and paste it to the assistant. The pack supports three workflow modes—`general`, `fast`, and `skill`—and reports benchmark gains (up to 76 % fewer generated lines while staying 100 % correct).

AIDA is an open‑source autonomous pentesting platform that lets any LLM (Claude, Gemini, OpenAI, etc.) drive a Docker‑based security‑testing environment. The agent can run reconnaissance, scanning, custom Python exploits, and raw HTTP manipulations, logging every command and its output in a persistent notebook. Findings are auto‑scored with CVSS 4.0 and can be exported as PDF reports. The project includes a web dashboard, JWT authentication, notification hooks, and supports swapping the built‑in `aida-pentest` container for the larger Exegol image. Real CVEs discovered by AIDA are listed, showing practical impact. Licensed under AGPL‑v3.

A mono‑repo of >100 Model Context Protocol (MCP) servers that expose cloud services, databases, dev tools, and more as natural‑language‑driven APIs for LLMs. Includes ready‑made servers for compute, storage, security, search, etc., with local/remote deployment, Python/TS SDKs, and MIT licensing.

Millet is an offline‑first Python tool that records dual‑channel audio from any meeting app, transcribes it with WhisperX (or MLX Whisper), diarises speakers via pyannote‑audio, and generates AI‑written summaries and professional PDFs. It offers three summarisation presets—high‑quality (cloud), confidential (TEE‑protected GLM‑5.3), and alternative (Kimi)—and supports multilingual transcription, voice‑print speaker recognition, structured markdown/YAML front‑matter, and optional Git sync. Linux is required for audio capture; macOS can run only the post‑capture steps.

A Microsoft‑maintained, MIT‑licensed tutorial series that teaches JavaScript developers to build generative‑AI apps. It covers LLM basics, prompt engineering, structured output, RAG, tool calling, and the Model Context Protocol, with code samples, quizzes, videos, and a companion chat app. You can start instantly in a GitHub Codespace or run the code locally.

Attention Gym is a PyTorch‑based library that supplies reusable mask/score‑mod functions, sparse‑attention patterns, and linear‑attention kernels for the FlexAttention API, plus example scripts and utilities for building and benchmarking custom attention mechanisms.

VRCT is a translation and transcription tool designed to help VRChat users communicate across different languages via real-time audio-to-text and chat integration.