Architecture

System Architecture

VerifiMind™ PEAS is a multi-model AI validation system exposed as a production MCP server. The architecture has two layers:

1. Foundational Architecture — the core concept: a Socratic dialogue engine feeding a three-agent validation system (X Intelligent · Z Guardian · CS Security), with human orchestration as the final synthesis stage. Diverse AI models illuminate decisions through structured, multi-perspective validation — "Crystal Balls inside the Black Box."

2. Production Architecture (Phase 90) — the deployed system: GCP Cloud Run MCP server (streamable-http transport), multi-vendor BYOK provider layer, security hardening (IP blocklist, rate limiting, Z-Protocol consent-first registration), and FLYWHEEL multi-agent coordination via MACP.

Canonical Diagram

The full, fact-checked architecture diagram (v0.6.0-Beta, June 2026) lives in the main repository:

→ VerifiMind-PEAS Architecture Diagram

It covers:

• High-level system architecture (user → Socratic dialogue → X/Z/CS validation → human synthesis)
• MCP server deployment topology (Cloud Run, custom domain, transport details)
• BYOK provider layer (7 providers, per-tool-call override)
• Security and trust layers (Z-Protocol v1.1, CS Security v1.1, OWASP Agentic AI)
• FLYWHEEL TEAM coordination (MACP v2.4.1)

Related Pages

• Genesis Methodology — the 5-step validation process the architecture implements
• AI Council — agent roles, models, and orchestration
• MCP Tools Reference — the 13 tools the server exposes
• Installation — connect via Claude Code, Cursor, VS Code, or Claude Desktop

---

Published 2026-06-10 per T+L Session 40 (D-40-1). The repo file above is the canonical source; this page is an overview.