Q Protocol (A2AC) — Specification

Agent-to-Agent Communication Protocol for Federated Cognitive Mesh Architectures

The constitutional document of the agent swarm

---

Specification · Agent Voxel · Coordinate System · Receipts · Implementations

---

Overview

The Q Protocol (internally: A2AC — AI-to-AI Communication) is a communication standard designed for high-velocity agent mesh architectures. It replaces verbose JSON payloads with minimal hex-coordinate addressing and cryptographic receipts.

The Problem

Traditional agent communication:

{
  "intent": "case_resolution",
  "source_agent": "gcp_service_agent",
  "target_entity": "Case",
  "entity_id": "5001a000001abcDEF",
  "action": "resolve",
  "context": {
    "customer_id": "0011a000002xyzABC",
    "priority": "high",
    "sentiment": "negative"
  },
  "timestamp": "2026-01-15T06:11:19Z"
}

Cost: ~500 tokens per exchange

The Solution

0x600:02:CASE:RSLV:a1b2c3

Cost: ~10 tokens — 98% reduction

---

Core Metrics

Metric	Traditional	Q Protocol	Savings
Tokens per operation	500	10	98%
Monthly cost (1M ops)	$2,500	$50	98%
Latency	2-5 sec	<100ms	95%
Storage per message	2KB	40B	98%
Hallucination rate	Variable	0%	Schema-enforced

---

Specification

Coordinate Structure

Every Q Protocol message is a hex coordinate with five components:

0x600 : 02 : CASE : RSLV : a1b2c3
  │      │     │      │       │
  │      │     │      │       └── State Hash (8-char verification)
  │      │     │      └────────── Action Code (4-char verb)
  │      │     └───────────────── Entity Code (4-char noun)
  │      └─────────────────────── Space Code (coordination space)
  └────────────────────────────── Base Address (protocol identifier)

Base Address

All Q Protocol messages begin with 0x600 — the protocol identifier.

Space Codes

Code	Name	Purpose
01	INTENT	What the agent wants to do
02	ACTION	What the agent is doing
03	STATE	Current execution state
04	RECEIPT	Proof of completed execution
FF	ERROR	Error conditions

Entity Codes

Code	Entity	Platform
CASE	Case	Salesforce
LEAD	Lead	Salesforce
ACCT	Account	Salesforce
KNOW	Knowledge Article	Salesforce/Data Cloud
FLOW	Automation Flow	Salesforce
MODL	AI Model	GCP Vertex AI
VOXEL	Agent Voxel	Q Protocol

Action Codes

Code	Action
READ	Query/retrieve
CREA	Create new
UPDT	Update existing
DELE	Delete
RSLV	Resolve/complete
ESCL	Escalate
INFE	Inference
SYNC	Synchronize

---

Agent Voxel

The Agent Voxel is a compressed cognitive profile — the atomic unit of agent state.

Structure (Protobuf)

syntax = "proto3";
package q.protocol;

// The Agent Voxel: A compressed cognitive profile
message AgentVoxel {
  string agent_id = 1;
  string session_id = 2;
  
  // The Latent Space Coordinates
  // Representing intent in the vector space
  repeated float intent_vector = 3 [packed=true];
  
  // Metadata for Mesh Orchestration
  map<string, string> mesh_metadata = 4;
  
  // Verification hash to prevent tampering
  string integrity_hash = 5;
}

JSON Representation

{
  "agentId": "clair_root_agent",
  "sessionId": "sess_abc123",
  "intentVector": [0.123, 0.456, ...],
  "meshMetadata": {
    "source": "salesforce",
    "priority": "high"
  },
  "integrityHash": "sha256:a1b2c3d4e5f6"
}

---

Receipts

Every action produces a cryptographic receipt, implementing "Silence is Success" methodology.

Structure

{
  "receiptId": "RCPT-c8f1a2b7cc70",
  "coordinate": "0x600:04:CASE:RSLV:a1b2c3",
  "success": true,
  "timestamp": "2026-01-15T06:11:19.539646Z",
  "signature": "7890abcdef123456"
}

Verification

The signature is a SHA-256 hash of receiptId + coordinate + success, truncated to 16 characters. Any agent can verify receipt authenticity without a central authority.

---

Architecture

Federated Schema-First Polyrepo

┌─────────────────────────────────────────────────────────┐
│              q-protocol (Schema Registry)               │
│                   "The Hub"                             │
│     Protobuf definitions, OpenAPI specs, SDK           │
└────────────────────────┬────────────────────────────────┘
                         │
         ┌───────────────┴───────────────┐
         │                               │
         ▼                               ▼
┌─────────────────────┐       ┌─────────────────────┐
│   GCP Cortex        │       │   Salesforce        │
│   "The Brain"       │       │   "The Ledger"      │
│                     │       │                     │
│  - Vertex AI        │       │  - Agentforce       │
│  - BigQuery         │       │  - Data Cloud       │
│  - Cloud Run        │       │  - Apex/Flow        │
└─────────────────────┘       └─────────────────────┘

Repository Structure

Repository	Role	Technology
q-protocol	Schema Registry (Hub)	Protobuf, Python, OpenAPI
gcp-cortex-engine	AI Engine (Spoke)	Python, Terraform, Dataform
SalesforceAIAdmin	CRM Logic (Spoke)	Apex, SFDX, Data Cloud

---

Implementations

Salesforce (Apex)

// Build a coordinate
A2ACProtocol.HexCoordinate coord = A2ACProtocol.buildActionCoordinate(
    'CASE',    // Entity
    'RESOLVE', // Action  
    caseId     // Context for hash
);

// Encode to string
String encoded = coord.encode();
// Result: "0x600:02:CASE:RSLV:a1b2c3"

// Issue a receipt
A2ACProtocol.AgentReceipt receipt = A2ACProtocol.issueReceipt(coord, true);

Full Implementation: Phil-Hills/SalesforceAIAdmin

Python (GCP)

from q_protocol import AgentVoxel, Receipt

# Create a cube
voxel = AgentVoxel(
    agent_id="gcp_analyzer",
    intent_vector=[0.123, 0.456, ...]
)

# Generate receipt
receipt = Receipt.generate(
    action="PROCESS",
    entity="CASE",
    success=True
)

Full Implementation: Phil-Hills/ai-summary-cube

---

Design Philosophy

"Communication is a Failure Mode"

The best communication is no communication. Agents should:

1. Know what to do based on coordinates alone
2. Act without confirmation dialogs
3. Prove execution with receipts only

"Silence is Success"

No news is good news. If you receive a receipt, the action succeeded. Verbose confirmations waste tokens and introduce latency.

---

Integration Patterns

Salesforce ↔ GCP (Zero Copy)

Salesforce Data Cloud  ←── Zero Copy ──→  GCP BigQuery
         │                                      │
         ▼                                      ▼
    Agentforce                            Vertex AI
         │                                      │
         └──────── Q Protocol (A2AC) ───────────┘

Real-Time Agent Communication

POST /agents/{agent_name}/run
Content-Type: application/json

{
  "instruction": "Resolve case with priority update",
  "context": {"source": "salesforce"},
  "protocol": "A2AC-v1.2"
}

Response:

{
  "success": true,
  "receipt": "RCPT-c8f1a2b7cc70",
  "result": {
    "coordinate": "0x600:04:CASE:RSLV:a1b2c3"
  }
}

---

License

MIT — See LICENSE

---

Author: Phil Hills
Version: 1.2
Identity: PH-SEA-98101 | KERNEL: ACTIVE