Tagentacle: The ROS of AI Agents 🐙

Tagentacle is a decentralized, configuration-centralized, production-grade multi-agent framework. It deeply adopts the software organization patterns of ROS 2 (Robot Operating System), combined with the modern AI ecosystem (MCP Protocol) and dynamic Schema technologies, to build a robust infrastructure for LLM multi-agent collaboration.

Everything is a Pkg. Managed. Verifiable. Scalable.

---

Quick Links

• 🚀 Getting Started — Install and run your first workspace
• 💡 Philosophy — Core design principles
• ⚔️ Why Tagentacle — Comparison with alternatives
• 🏗️ Architecture — System overview
• 🤖 Agent Architecture — IO + Inference separation
• 🐍 Python SDK — Dual-layer API
• 🔌 MCP Integration — Local sessions + Direct HTTP
• 🔐 TACL — JWT access control
• 📋 Best Practices — Recommended patterns
• 🐳 Containers — Container-ready deployment

The Architecture Stack

┌─────────────────────────────────────────────────────┐
│  Application Layer (Nodes, Agents, MCP Servers)      │  ← iterates fastest
├─────────────────────────────────────────────────────┤
│  Tagentacle (Domain Shell / Middleware)               │  ← domain semantics
│  Topic Pub/Sub, Service RPC, /mcp/directory,         │     (syntax sugar with
│  Node identity, Lifecycle, TACL, Launch              │      real engineering value)
├─────────────────────────────────────────────────────┤
│  Linux (Kernel + Docker)                             │  ← "laws of physics"
│  socket, namespace, cgroup, filesystem, signals      │
└─────────────────────────────────────────────────────┘