Orkes

Orkes helps you coordinate LLM agents using plain Python. No magic, no unnecessary layers. Just explicit control flow, transparent agent logic, and simple message passing.

The goal of this effort is to make my trained model easily usable and inference-ready for anyone. I believe the potential applications of this work span across various domains, and I am excited to see how others might extend and build upon it.

🔹 Core Principles

• Explicit control flow — use DAGs, FSMs, or plain loops
• Transparent agents — define prompt, tool, and logic directly
• Simple message passing — plain dicts, no graph state magic
• Minimal dependencies — only what you truly need
• 100% Pythonic — easy to read, modify, and extend
• Stateless by default — you control memory and state
• Hackable and debuggable — nothing hidden

👉 Check out the repo