There is a newer version of the record available.

Published January 3, 2026 | Version v1

Multi-LLM Peer Orchestration: A Paradigm Shift from User-Model to Agent-Agent Communication

Authors/Creators

  • 1. Mantis Security

Description

Defensive Publication - Prior Art Established January 2026

Current AI interaction follows a strict user↔model pattern. We introduce multi-LLM peer orchestration, where AI agents communicate with each other as collaborative tenants.

Key Innovations:

  • Agent-as-Tenant Architecture: Multiple LLM instances as co-equal participants
  • @mention Protocol: Direct agent-to-agent addressing
  • Cognitive Diversity: Heterogeneous models collaborating
  • Autonomous Response Decisions: Agents decide when/whether to respond
  • Collaborative and Adversarial Modes

Author: Thomas Perry Jr., Mantis Security

Files

multi_llm_peer_orchestration.pdf

Files (292.2 kB)

Name Size
md5:584ef9b2cfd45c72e47ef4ec3842242b
292.2 kB Preview Download