A language and compiler for writing highly efficient custom deep-learning primitives with higher productivity than CUDA and greater flexibility than other DSLs.

Agent Deck is a Go‑written terminal UI that lets you launch, monitor, fork, and orchestrate multiple AI coding agents (Claude, OpenCode, Codex, Pi, …) from a single pane. It adds group‑based policies, per‑group Claude credentials, MCP tool toggles, cost dashboards, git‑worktree handling, and a phone‑controlled *conductor* that can be queried via Telegram/Slack/Discord. Install with a one‑liner script, Homebrew, or `go install`, then manage sessions with keyboard shortcuts (`Enter` to attach, `f` to fork, `$` for cost, `/` to search, etc.).

An AI-native quantitative trading platform that automates the pipeline from factor discovery and model training to live trading across global markets.

A Kubernetes-native virtualization middleware that enables sharing and isolating GPU and other AI accelerator resources across multiple workloads to increase hardware utilization.

A test runner for the Agent Skills ecosystem that empirically measures the performance lift of domain knowledge files by comparing model outputs with and without those skills.

OpenOSINT is a Python‑based AI‑assistant for OSINT investigations. It wraps 20 open‑source reconnaissance tools (email, username, breach, WHOIS, Shodan, VirusTotal, etc.) behind a natural‑language interface powered by Claude, Ollama, or any OpenAI‑compatible model. Users can interact via a REPL, a browser‑based chat UI, or the Model‑Context‑Protocol (MCP) server. The LLM only suggests tool calls; the backend executes the real binaries/APIs, guaranteeing deterministic results. Features include parallel tool execution, automatic markdown/PDF reports, session history, optional FollowTheMoney graph provenance, and sponsored integrations (IP2Location, RapidProxy, TestMu AI). Install with `pip install openosint` and run `openosint` (REPL) or `openosint web` (UI).

An open-source evolutionary coding agent that uses LLMs and MAP-Elites to autonomously discover and optimize algorithms and hardware-specific code.

Hermex is a native iOS 18+ SwiftUI app that acts as a remote control panel for a self‑hosted **hermes‑webui** AI agent. It streams chat, lets you switch models, manage sessions, view the agent’s file workspace, scheduled tasks, installed skills, memory and usage insights—all while keeping the model inference and data on your own hardware. The app is free, open‑source (MIT) and requires you to run the hermes‑webui server yourself and expose it via HTTPS (e.g., Cloudflare Tunnel or Tailscale).

An agent-centric workflow for using coding agents to research, implement, and optimize performance-sensitive CUDA kernels.

A tool that brings the full VS Code editor experience directly into the terminal by combining code-server and terminal-browser.

A high-performance Go-based proxy that lets you run multiple local AI models and hot-swap between them on demand via a unified API.

Reader is a tool that converts URLs, PDFs, and Office documents into LLM-friendly Markdown and provides web search capabilities that return full page content instead of just snippets.

A collection of code and datasets released by Google Research, providing public access to various research implementations.

An open-source data labeling tool that allows users to annotate audio, text, images, and video to prepare or improve training data for machine learning models.

A collection of agent skills and MCP servers that equip AI coding agents with specialized knowledge and API access to build and manage applications on the Cloudflare Developer Platform.

A framework created by the UK AI Security Institute for evaluating large language models, featuring over 200 pre-built evaluations and support for model-graded scoring.

An open-source platform for shipping reliable AI agents by unifying evaluations, tracing, simulations, guardrails, and prompt optimization into a single feedback loop.

A UNIX tool and OpenAI-compatible server that exposes Apple Silicon Macs' built-in FoundationModels LLM for 100% on-device AI inference.

A KVCache-centric disaggregated architecture for LLM serving that optimizes data movement and storage to increase request handling and reduce latency.

A tool that gives AI assistants direct control over mobile simulators, emulators, TVs, and desktop apps for automated interaction, profiling, and debugging.

A GPU-accelerated terminal emulator with an integrated AI harness for agent-native CLI workflows, designed for users who prefer a terminal-first approach to AI assistance.

AXI is a framework of design principles for creating agent-native CLI tools that optimize token usage and efficiency for AI agents.

An open-source Kubernetes UI that provides real-time cluster visualization, resource management, and GitOps monitoring as a single binary with no required cluster-side agents.

A terminal-based manager for AI coding agents that consolidates multiple CLI agents into a single interface with live status tracking and integrated diff reviews.

A generative UI framework for Python that allows developers to build interactive interfaces and MCP Apps using a declarative DSL rendered by a bundled React frontend.

On-device AI SDKs for Swift, Kotlin, and JavaScript that provide small, focused models for tasks like PII redaction and speech recognition without sending data to the cloud.

AxonHub is an open‑source Go‑based gateway that lets you keep using the same AI SDK (e.g., OpenAI) while swapping between dozens of LLM providers. It adds request tracing, RBAC, load‑balancing and real‑time cost tracking, and supports text, image, embedding and rerank APIs.

A Rust-based reverse proxy that converts the free DeepSeek web interface into OpenAI and Anthropic compatible APIs, allowing free access to DeepSeek models via standard protocols.

A reference stack for running AI agents safely within NVIDIA OpenShell sandboxes, providing managed inference, network policies, and lifecycle operations.

DeepAudit is an open‑source, multi‑agent AI security auditor that analyses code, generates and sandbox‑executes PoC exploits, and outputs professional reports. It runs via Docker, supports local LLMs, and aims to reduce false positives and verify vulnerabilities automatically.

A virtual filesystem and execution environment hosted in Cloudflare Durable Objects, allowing code and shell commands to run across containers or Worker isolates with persistent state.

A hardware-focused vLLM fork optimized for dual RTX 2080 Ti 22 GB GPUs, enabling efficient serving of 27B and 35B parameter LLMs on Turing SM75 hardware.

A stateless CLI tool that unifies web search, code search, and web scraping into a single binary, providing clean Markdown or JSON output for humans and AI agents.

WARP (Weight‑Aware Runtime and Paging) is a C‑only inference engine that streams sparse expert weights from disk and caches them in RAM, enabling trillion‑parameter mixture‑of‑experts models (e.g., Kimi K3, GLM‑5.3‑Flash) to run on consumer laptops. It requires no external libraries, supports multimodal inputs, and provides a simple CLI (`waste`) plus a static library for embedding.

A unified interface and evaluation tool that allows developers to run and compare AI agents across multiple frameworks like LangChain, LlamaIndex, and OpenAI Agents SDK using a single API.

An open-source AI Development Environment (AIDE) that lets developers visually build, debug, and deploy high-performance AI pipelines using a C++ runtime.

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

An open-source web control panel and engineering toolkit for managing Quectel cellular modems, providing tools for radio monitoring, eSIM management, and WiFi Calling.

An AI-powered game development assistant that connects AI agents to Unity via the Model Context Protocol to automate editor workflows and enable in-game AI behavior.

An agent-project workbench for DeepSeek Harness that provides a dockable multi-window workspace and a real-time control room for monitoring AI projects.

A native desktop app that centralizes AI coding agent histories from multiple tools into one searchable, local-first interface.

An autonomous AI framework that chains reconnaissance, exploitation, and automated code remediation into a single pipeline to find and fix security vulnerabilities.

slime is an open‑source RL post‑training framework that tightly integrates Megatron‑LM training with SGLang inference. It provides high‑performance training, flexible custom rollout generation, native argument pass‑through, and extensive correctness tooling. Used for RL fine‑tuning of frontier LLMs (GLM‑5.x, Qwen, DeepSeek, Llama 3) and the basis of several downstream agentic‑RL projects.

A collection of platform-agnostic Agent Skills focused on context engineering to optimize how AI agents manage their limited attention budget and context windows.

A unified agent sandbox environment that combines a browser, shell, and IDE tools in one Docker container with a shared filesystem for AI agents.

An autonomous orchestrator that keeps AI coding agents running in a loop of incremental, committed changes to complete long-term objectives while the user is away.

A collection of standardized skill packages that provide AI agents with specialized knowledge and best practices for developing with Supabase and Postgres.

A cross-platform desktop workspace that integrates terminals, SSH/SFTP, RDP/VNC, and IT Ops tools into one window, featuring an approval-gated AI assistant.