Sentinel MCP — AI-Orchestrated Recon Platform
Builder: Nirmal Liyon, VertexElite Security Research Period: April–June 2026 Status: Internal AI infrastructure (not standalone product)
Architecture
30-tool AI-orchestrated recon platform with 6-phase pipeline.
Specialist Model Routing
| Model | Role |
|---|---|
| cogito-2.1:671b | Vulnerability logic analysis |
| deepseek-v3.2 | Reconnaissance operations |
| kimi-k2:1t | Threat modeling |
| kimi-k2-thinking | Attack chain analysis |
| gemini-3-flash-preview | Sanitization/trust-boundary layer between CN models and Claude |
Trust Architecture
- gemini-3-flash-preview served as sanitization layer between Chinese-origin models and Claude
- Retired from Ollama and replaced with Google AI Studio API direct integration
- Ensures no direct data flow from untrusted CN models to Claude context
Classification
MCPs are internal AI infrastructure, not standalone products. Part of VertexElite's operational security research toolkit.
Published by VertexElite Security Research · research@vertexelite.org