AI Council

AI Council: Multi-Model Orchestration

Agent roles, models, and orchestration architecture

---

Overview

The AI Council is the core orchestration layer of VerifiMind™ PEAS. It synergizes diverse AI perspectives under human direction for objective, validated results. The human orchestrator sits at the center, directing all AI agents and making final decisions.

---

The Four Agents

Y — The Innovator

Property	Value
Role	Creative Strategy
Default Model	Gemini 2.5 Flash
MCP Tool	consult_agent_x (shared tool, agent selection via parameters)
Strength	Generates creative concepts, strategic insights, and novel approaches
Added	v0.4.3

The Innovator brings fresh perspectives and creative solutions. It excels at brainstorming, strategic planning, and identifying opportunities that analytical agents might overlook.

X — The Analyst

Property	Value
Role	Critical Analysis
Default Model	Gemini 2.5 Flash
MCP Tool	consult_agent_x
Strength	Provides critical analysis, identifies weaknesses, and stress-tests assumptions
Since	v0.1.0

The Analyst is the skeptic of the council. It challenges every claim, identifies logical gaps, and ensures that conclusions are supported by evidence.

Z — The Guardian

Property	Value
Role	Ethics & Safety
Default Model	Gemini 2.5 Flash
MCP Tool	consult_agent_z
Strength	Ensures ethical compliance, Z-Protocol enforcement, veto power
Since	v0.1.0

The Guardian holds veto power over any recommendation that violates ethical principles. It enforces the Z-Protocol — a structured framework for ethical AI decision-making.

CS — The Validator

Property	Value
Role	External Validation
Default Model	Gemini 2.5 Flash
MCP Tool	consult_agent_cs
Strength	Validates claims against external evidence, 4-Stage Security Verification
Since	v0.2.0

The Validator ensures that all claims are grounded in verifiable facts. It runs the 4-Stage Security Verification Protocol for security-related findings.

---

BYOK Provider Override (v0.4.5+)

With BYOK, users can override any agent's provider on a per-tool-call basis:

Use consult_agent_x with api_key "gsk_your_key" to analyze the market.

This creates an ephemeral provider instance — your key is used once and immediately discarded. See the BYOK Guide for full details.

Supported Providers

Provider	llm_provider	Auto-Detect
Gemini	(default)	N/A
OpenAI	openai	sk-
Anthropic	anthropic	sk-ant-
Groq	groq	gsk_
Mistral	mistral	Manual
Ollama	ollama	Manual
xAI	xai	Manual

---

Full Trinity Pipeline

The run_full_trinity tool executes all four agents in sequence:

1. Y (Innovator) generates initial concepts
2. X (Analyst) critically examines the concepts
3. Z (Guardian) reviews ethical implications
4. CS (Validator) validates against external evidence
5. Human receives the synthesized multi-perspective report

---

Architecture Diagram

                    ┌─────────────────┐
                    │  Human Director  │
                    │  (Orchestrator)  │
                    └────────┬────────┘
                             │
              ┌──────────────┼──────────────┐
              │              │              │
        ┌─────┴─────┐ ┌─────┴─────┐ ┌─────┴─────┐
        │ Y:Innovate │ │ X:Analyze │ │ Z:Guard   │
        │ (Gemini)   │ │ (Gemini)  │ │ (Gemini)  │
        └─────┬─────┘ └─────┬─────┘ └─────┬─────┘
              │              │              │
              └──────────────┼──────────────┘
                             │
                    ┌────────┴────────┐
                    │  CS: Validate   │
                    │  (Gemini)       │
                    └─────────────────┘

With BYOK, any agent's model can be swapped at call time without changing the architecture.

---

← Back to Home