Gliding Horse Agent OS: An Open-Source AI Agent Learning Platform

Forum topic · doiito · 2026-06-05

Summary

Gliding Horse is a fully open-source AI agent operating system developed by a member of the zhichai.net community and shared as a collaborative learning platform. Hosted on GitHub (doiito/gliding_horse), the project orchestrates multiple AI agents through a PDCA (Plan-Do-Check-Act) loop, positioning itself as an infrastructure layer—a 'harness'—that organizes agents into collaborative, auditable, and self-evolving systems. The core orchestration layer is written in Rust, featuring a 5W2H ontology, an event bus, and lifecycle management. A five-tier hierarchical memory system combines Sled, Qdrant, and Oxigraph with MESI consistency, while a data bus based on JSON-LD 1.1 provides interoperability via @id/@type/@context and named graphs. Knowledge and skill graphs are stored in RDF with SPARQL 1.1 support; the skill graph spans 7,500 lines and is designed for self-evolution. A perception engine offers 10 trigger types with anomaly deduplication, and the gateway exposes gRPC, HTTP (OpenAI-compatible), and MCP interfaces. The project draws metaphorical inspiration from Zhuge Liang's legendary 'Wooden Ox and Gliding Horse' transport devices of the Three Kingdoms era, emphasizing adaptive orchestration over rigid frameworks. Design details are documented in the repository for community study.

The developer of Gliding Horse invites community members interested in AI agent development to learn and discuss together. The project is fully open source, including detailed design documentation: https://github.com/doiito/gliding_horse

What is Gliding Horse?

Gliding Horse is an AI agent operating system that coordinates multiple AI agents through a PDCA (Plan-Do-Check-Act) loop. It serves as an infrastructure layer—a harness for AI agents—organizing them into collaborative, auditable, and self-evolving systems, likened to Zhuge Liang's legendary wooden oxen and gliding horses that revolutionized logistics on treacherous mountain roads.

> "We don't just build agents; we build the infrastructure to harness collective intelligence."

Core Technology Stack

| Layer | Technology | Responsibility | |---|---|---| | Core Orchestration (Rust) | PDCA loop · 5W2H ontology · Event bus | Agent orchestration and lifecycle management | | Memory System | L0: Sled+Qdrant · L2: Oxigraph · MESI consistency | Five-tier hierarchical memory | | Data Bus | JSON-LD 1.1 · @id/@type/@context · Named graphs | Universal interoperability layer | | Knowledge Graph | Oxigraph RDF · SPARQL 1.1 · Code AST | Unified storage across subsystems | | Skill Graph | RDF · 7,500 lines · Self-evolving | Dynamic cognitive network | | Perception Engine | 10 trigger types · Anomaly deduplication · 5W2H constraint checks | Proactive monitoring | | Gateway | gRPC · HTTP (OpenAI-compatible) · MCP | Production-grade interfaces |

The Story: From Ancient Wisdom to Modern Intelligence

During the Three Kingdoms period (220–280 AD), the legendary strategist Zhuge Liang (Chancellor of Shu Han) faced a severe challenge: how to transport supplies efficiently through Sichuan's treacherous mountain roads during northern campaigns. Traditional wheeled carts struggled on narrow, steep paths, and human porters were quickly exhausted by limited carrying capacity.

His solution—the Wooden Ox and Gliding Horse (木牛流马)—were autonomous transport devices that could traverse complex terrain with minimal human guidance. They represented a paradigm shift: autonomous systems extending human capability.

Bridging Past and Present: The Agent Harness

Just as the gliding horse served as an intelligent harness for transporting supplies across impassable terrain, Gliding Horse Agent OS acts as an intelligent harness layer for AI agents:

| Ancient Innovation | Modern Implementation | |---|---| | Autonomous transport | Self-driven agent workflows | | Terrain adaptation | Dynamic complexity handling (7 levels) | | Load distribution | Parallel agent execution | | Minimal guidance | Proactive anomaly detection | | Mechanical reliability | Rust memory safety |

> "The skilled warrior adapts to circumstances and turns them to advantage, like pouring water into water." > — Zhuge Liang

This ancient wisdom guides the design philosophy: flexible orchestration that adapts to task complexity, rather than rigid frameworks that force tasks into predetermined molds.

Interested developers are welcome to join the discussion and learn together via the GitHub repository.