Explore the directory

A shared intelligence runtime for AI agents that gives them persistent, multi-type memory, compressed web context and cross-agent evolution, so one agent's fix becomes another's recommendation.

• Agent memory

An open-source API gateway that routes, manages and analyzes requests to multiple LLM providers behind one OpenAI-compatible interface, hosted or self-hosted with Docker.

• LLM gateways

A pre-execution guard that blocks destructive Git and filesystem commands and access to secrets such as SSH keys before a coding agent's tool call actually runs.

• Coding agents
• Security

A CLI and set of standards for starting a new agent project, scaffolding scenario tests, versioned prompts and evaluation notebooks so agent behavior stays testable from day one.

• Agent frameworks
• AI evals

A native, editable design system for DeepSeek Harness that lets the AI generate and continuously revise real websites, app prototypes, posters and slide decks as saved project files.

• Design tools

A self-hosted CLI agent framework that automates development chores such as PR reviews, bug fixing and security patching by combining reusable steps, prompt templates and an LLM.

• Agent frameworks
• Code quality

A Windows-first native desktop client for Hermes Agent's Chinese-community core, built with Tauri, Rust and React, adding a local window shell around the existing agent dashboard.

• Coding agents

The official MATLAB Model Context Protocol server from MathWorks, letting AI applications such as Claude Code and VS Code Copilot start MATLAB, write and run code, and check style.

• MCP servers

A TensorFlow toolkit that optimizes Keras and TensorFlow models for deployment through techniques such as quantization and pruning for sparse weights.

• Models

An Electron desktop app that arranges AI coding agents, terminals, tasks and notes on one infinite 2D canvas so a developer can see and resume parallel work at a glance.

• Coding agents

A PyTorch research library for neural network quantization, supporting both post-training quantization and quantization-aware training for common layer types.

• Models

A high-performance inference engine for LLM, VLM, DiT and REC models, optimized for Chinese AI accelerators and donated to the OpenAtom Foundation.

• Inference

Shared project memory for engineers and their coding agents, stored as reviewable Markdown in the repository itself and synced through ordinary Git commits.

• Agent memory

A method-native AI work system and terminal CLI for coding that routes across GPT, Claude, DeepSeek, Qwen, Kimi, GLM and MiniMax, with persistent, resumable sessions and governed tools.

• Coding agents
• LLM gateways

An enterprise security system, deployed in production at Uber, that discovers, observes, benchmarks and detects risky behavior in employee-facing and customer-facing AI agents.

• Security

The official MiniMax Model Context Protocol server that lets Claude Desktop, Cursor, Windsurf and other MCP clients generate speech, clone voices, and generate images and video.

• MCP servers
• Speech
• Generative media

A Docker CLI plugin and gateway that runs Model Context Protocol servers as isolated containers, giving every AI client in an organization one consistent, secured connection point.

• MCP servers
• Containers

A production-grade MCP server giving an AI model twenty-eight security-intelligence tools across two dozen sources, including NVD, EPSS, CISA KEV, Shodan and VirusTotal, correlated behind one call.

• MCP servers
• Security

A memory operating system for personalized AI agents, using an operating-system-style hierarchical storage architecture to keep long-term context coherent across conversations.

• Agent memory

A native local-development system for Apple Silicon combining Docker, Compose, one-click Kubernetes, full graphical Linux virtual machines and policy-bound coding-agent sandboxes, without requiring Docker Desktop.

• Containers

A Model Context Protocol server that lets an AI model drive WinDbg for crash-dump triage, live user-mode debugging and kernel debugging using natural-language requests.

• MCP servers
• Debugging

A Model Context Protocol server that connects to a Kubernetes cluster and lets an AI agent manage it through kubectl and Helm, loading credentials from the usual kubeconfig sources.

• MCP servers
• Containers

A vLLM-style inference server for Apple Silicon that adds continuous batching, paged and prefix KV caching, and both OpenAI and Anthropic-compatible APIs in one process.

• Inference

A ready-to-fork Claude Code template for academic work in LaTeX, Beamer and R, with specialized review agents, quality gates and replication checks.

• Coding agents