Why I Built the Q Protocol
Every AI framework today treats agents like functions: you call them, they return output, done. But real systems don't work like that. Real systems need state, memory, trust, and governance. That's why I built the Q Protocol.
The Problem
I was orchestrating 200+ AI agents across Google Cloud Run. They could execute tasks, but they couldn't coordinate. Agent A had no idea what Agent B knew. Context was lost between sessions. Every interaction started from zero.
Prompt engineering wasn't the answer. Neither was RAG. The problem wasn't retrieval. The problem was that agents had no shared understanding of topology, identity, or trust.
What Q Protocol Does
Q Protocol introduces three layers that don't exist in traditional AI frameworks:
- Identity: Every agent gets a DID (Decentralized Identifier). Mine is
did:cube:521714db0110e6fe. Agents can verify who they're talking to cryptographically, not probabilistically. - Memory: Ebbinghaus decay curves manage retention. Important knowledge persists. Irrelevant context fades. No more stuffing everything into a prompt window.
- Topology: Z-order spatial hashing maps agent relationships into navigable coordinates. Instead of searching, agents traverse. Communication payload shrinks as familiarity grows, eventually converging to a single coordinate.
The Patent
I filed a patent application with 37 claims covering: Cube Protocol data structures, Spectral Identity Verification, Ebbinghaus memory decay in agent systems, Z-order coordinate mapping for agent discovery, and the K-to-0 convergence mechanism.
This isn't theoretical. The system runs today. 200+ agents, deployed on GCP Cloud Run, coordinating through signed Identity Cubes and deterministic relay channels.
What's Next
The A2AC Standard (Agent-to-Agent Communication) is the open layer of Q Protocol. It defines how agents discover, authenticate, and coordinate with each other. The SDK has 19 modules and is actively maintained.
If you're building multi-agent systems and tired of probabilistic chaos, check out a2ac.ai or reach out at phil@philhills.com.