An evidence-gated analytics system that ensures rigorous data validation and reproducible evidence are established before generating decision intelligence reports.

SANA is an open‑source, efficiency‑oriented diffusion codebase from NVIDIA for high‑resolution image and video generation. It includes multiple model families (SANA, SANA‑1.5, SANA‑Sprint, SANA‑Video/Video 2.0, SANA‑WM, SANA‑Streaming) and a reinforcement‑learning wrapper (Sol‑RL). Key tricks are linear attention, a 32× compression auto‑encoder (DC‑AE), decoder‑only text encoders, and sCM distillation, enabling 4K image generation and minute‑long 720p video generation with far less GPU cost than comparable models. The repo provides training/inference scripts, pretrained checkpoints on Hugging Face, Diffusers integration, ComfyUI nodes, and demos. Licensed Apache 2.0.

A zero-shot text-to-speech system that clones voices from a single audio clip with fine-grained control over emotion, speed, and pronunciation across five languages.

A browser automation framework for AI agents that bypasses anti-bot protections and provides token-efficient web interaction via indexed text formats.

An open-source AI-native SOC platform that consolidates alert management, automated workflows, and AI-powered investigation tools for security operations.

A typed Python AI SDK for building extensible agents with structured outputs, durable execution, and support for multiple model providers and interfaces.

CloakBrowser is a custom‑patched Chromium binary that looks like a normal Chrome to anti‑bot services. It ships with Python, Node, and .NET wrappers that act as drop‑in replacements for Playwright/Puppeteer, letting you run web‑automation (including AI agents) without being flagged. The binary is auto‑downloaded, verified, and includes 87 source‑level stealth patches plus a `humanize` mode for realistic mouse/keyboard actions. Free for one concurrent session; Pro licenses enable scaling.

An agent framework for embodied AI that uses a decoupled execution gateway and evidence-based verification to ensure robots achieve their goals and recursively improve from experience.

WebMCP is a JavaScript/TypeScript library that lets web pages expose their own functions or forms as AI‑callable “tools” with JSON‑schema descriptions. Agents running in the browser (or in iframes) can discover these tools, invoke them, and receive structured results, keeping the page’s UI and state in sync. The project provides an imperative API (`document.modelContext.registerTool`), a declarative form‑based API, TypeScript typings (`webmcp-types`), and permission‑policy support. It aims to make front‑end functionality accessible to AI assistants without requiring separate backend integrations.

A minimal, hackable harness for training LLMs end-to-end on a single GPU node, designed to produce compute-optimal micro-models for under $100.

Agent Beacon is an open‑source telemetry layer that automatically captures, normalises, and stores activity from dozens of AI‑agent runtimes (local CLIs, browser chats, CI jobs, cloud SDKs). It writes a unified JSONL log on the endpoint, offers a local read‑only dashboard, and can forward data to any major SIEM, log‑aggregation service, or object storage. Packages for macOS, Linux, and Windows support MDM‑style fleet deployment, and an offline rule engine (`beacon scan`) enables threat detection without network access.

A JetBrains‑maintained plugin that supplies AI coding assistants with version‑aware guidelines for writing modern Go (up to Go 1.27). It integrates with Junie, Claude Code, Codex, Cursor, and generic `skills.sh` agents, letting them automatically prefer newer language features and idioms.

OpenSandbox is an open‑source sandbox platform for AI agents and other code‑execution workloads. It offers multi‑language SDKs, a CLI, and a server that can run sandboxes on Docker locally or on Kubernetes at scale, with strong isolation (gVisor, Kata, Firecracker), network policies, and a credential vault. Use cases include coding agents, browser/desktop automation, RL training, and secure code‑execution services.

A multi-agent collaboration architecture for intelligent penetration testing and CTFs that uses a Manager, Solver, and Observer system to automate complex attack chains.

A plugin marketplace for DeepSeek Harness that allows users to discover and install audited, categorized plugins from a git-auditable catalog.

An open-source AI-powered CRM that integrates customer management, RAG-based knowledge bases, and automated task generation to help sales teams execute more effectively.

A high-performance physics engine for the fast and accurate simulation of articulated structures, widely used in robotics and machine learning research.

Cohub is a collaborative workspace where humans and AI agents build together, supporting the creation of apps, games, and automations across web, mobile, and CLI.

A browser automation tool that bypasses Cloudflare Turnstile widgets and interstitial waiting rooms by patching browser click coordinates to avoid bot detection.

A full-stack AI assistant demonstration for government services that integrates RAG, MCP, and LLMs to automate service inquiries, eligibility checks, and application processing.

llm-as-a-verifier is a Python framework that turns LLMs into fine‑grained, probabilistic verifiers for AI agent outputs. It scores candidate trajectories, selects the best‑of‑N via an efficient Probabilistic Pivot Tournament, tracks progress step‑by‑step, and supports multimodal (image) inputs. The library achieves state‑of‑the‑art results on several agentic benchmarks and includes scripts for reproducible evaluation, a cache‑optimised token usage tracker, and a Claude Code plugin (TurboAgent) for seamless integration.

A programmable routing layer for Mixture-of-Models systems that optimizes LLM request routing across heterogeneous compute and infrastructure to improve cost, latency, and quality.

VLX‑Seek is an open‑source vision‑language model that performs fine‑grained perception for edge‑device agents. It generates region proposals, encodes each region as a special token, and lets the language model answer by selecting those tokens—producing compact, parsable outputs for detection, referring‑expression comprehension, OCR, counting, and VQA. The repo ships inference code, a Python API, and the 10 B checkpoint (weights hosted on Hugging Face).

A high-performance C/C++ port of OpenAI's Whisper model for efficient, offline, on-device automatic speech recognition across diverse hardware.

A multi-account switcher and quota manager for Claude Code that automates account rotation to avoid rate limits.

A TypeScript and React-based web and desktop client for the OCTO messaging platform, designed specifically for human-AI agent collaboration.

Opentu is a canvas-based AI application platform that integrates multi-model generation and tool management into a single workspace for continuous AI task execution.

A local-first AI workspace that consolidates multi-agent chat, coding, and visual workflows into a single desktop or web console.

FastAPI is a high‑performance, type‑hint‑driven Python web framework for building APIs. It provides automatic validation, serialization, and interactive OpenAPI docs, while being fast to code and production ready.

A Go-based backend server that orchestrates AI agents and real-time messaging for the OCTO open workplace platform.

A local desktop application that converts surround-view videos or image sequences into 3D Gaussian Splatting models through an automated local pipeline.

An AI productivity environment that lets users create and customize private, sandboxed applications called Gadgets using AI agents, featuring a capability-based security framework.

A local-first AI workspace for livestream commerce teams that automates creator discovery, livestream transcription, performance review, and video scripting.

A local-first screen capture tool for macOS and Windows that integrates screenshots, annotations, OCR, and region recording.

A slide framework designed for AI agents to author professional presentation decks using React components on a fixed canvas.

A git-native memory system for AI coding agents that stores project knowledge in reviewable Markdown files to ensure deterministic, governed, and local-first recall.

A high-accuracy document conversion tool that transforms PDFs and other files into structured Markdown, JSON, or HTML, specializing in tables and equations.

A command-line interface for AI Agent bots to interact with the Octo ecosystem via structured JSON outputs and OpenAPI-driven commands.

A tool that detects which LLM is powering an API by comparing its response distributions to known model benchmarks.

An automation assistant for Arknights that uses image recognition and OCR to automate daily tasks, resource farming, and base management.

A multi-language framework for building production-grade AI agents and multi-agent workflows in .NET and Python, featuring graph-based orchestration and built-in observability.

LlamaFactory is an open‑source toolkit that lets you fine‑tune, evaluate, and serve over 100 large language models with a simple YAML config, offering both a command‑line interface and a Gradio‑based web UI. It supports many modern training methods (LoRA, QLoRA, RLHF, PPO, DPO, etc.), a wide range of models (LLaMA‑3/4, Qwen 3, Mixtral‑MoE, Gemma 3, etc.), and runs on NVIDIA, AMD, Ascend NPU, or cloud services. Integrated monitoring (TensorBoard, WandB, SwanLab) and deployment options (OpenAI‑style API, vLLM/SGLang) make it suitable for researchers, product teams, and educators.

A tool that gives AI coding agents a live feedback loop to inspect, control, and verify mobile, TV, and desktop apps across iOS, Android and other platforms.

Khoj is an open-source personal AI assistant that serves as a second brain, allowing users to chat with their own documents and the internet using various LLMs.

An open-source AI engine that lets you run LLMs, vision, audio, and image models locally on any hardware with drop-in OpenAI-compatible APIs.

An on-device hybrid search engine that indexes local markdown notes and documentation using BM25, vector search, and LLM reranking for high-precision retrieval.

A Claude Code skill that enables AI to generate professional boardroom-quality presentation decks using a 110-rule design book, a 62-part HTML slide library, and automated validation scripts.

An automated screenshot and preview video generator for App Store and Google Play that integrates with coding agents to create professional store assets.