Equitus Fusion Model Context Protocol (MCP): server controls cost and risk by acting as a choke point.

 

“Fusion MCP connects AI-assisted reasoning to ARCXA’s mapping and validation workflows—helping teams draft mappings before data moves, resolve uncertainty with evidence, and reuse approved integration knowledge.”

Equitus Fusion and Arcxa combine; Triple Store architecture (Subject–Predicate–Object), Knowledge Graphs, and Knowledge Graph Neural Networks (KGNNs) fundamentally transform how an LLM interacts with enterprise data and physical security infrastructure. 

MCP - Creates a bridge to combine different work flow layers; the LLM Layer which does the reasoning , ArcXA Layer does the touching of data, while MCP Layer as the boundary between them. 

An LLM is good at reasoning across systems but shouldn’t hold credentials, touch raw data, or act on its own authority. The MCP server lets it do the reasoning while the Equitus stack keeps the data, identity, and enforcement.

Pre-mapping becomes a conversation between the model and a set of read-only, metadata-level tools, so nothing physical moves until a mapping has been proposed, validated, and approved.

Equitus Fusion MCP (Model Context Protocol) Server serves as a secure, standardized middleware layer that standardizes how Large Language Models (LLMs) interface with Equitus Fusion, Arcxa, EVS (Equitus Video Sentinel), and enterprise ICAM (Identity, Credential, and Access Management) systems.

________________________________________________________________

MCP enables LLM agents to act as autonomous operational co-pilots capable of data migration, cross-domain integration, and unified cyber/physical security orchestration.

"Enterprise security and migration work is fragmented"

Video analytics, access control, identity systems, legacy databases, and alert pipelines each have their own schema, API, and permission model, and the connections between them are manual. 

Equitus Fusion Model Context Protocol (MCP): server controls cost and risk by acting as a choke point. The LLM never touches data, credentials, or infrastructure directly. It makes narrow, metered, policy-checked calls through the server, and the heavy work runs on infrastructure you control.

Plugging an LLM directly into each one would mean custom integrations per system, broad credentials sitting near a model, and no consistent audit trail.

MCP server then replaces all of that with one standard interface, so the model calls named tools and the platform decides what each call is allowed to do.

______________________________________________________

Equitus Fusion MCP Orchestrates the Ecosystem: Providing an MCP layer that works with network to connect Cyber/Physical Security Layer. Connecting Equitus Video Sentinel, Identity Control Access Management with Cyber/Physical access allows Power11 users to improve compliance and safety costs using intelligent video layers.

MCP systems, Provides multi-layered support for network safe security combining:  access management, combining video feeds, personal records, access passwords and semantic "purpose" tests will track  identity for cyber or physical access.  

1. Connecting to ICAM (Identity, Credential, and Access Management)

Before executing queries or physical/cyber controls, the MCP server enforces strict authorization through enterprise ICAM protocols (such as SAML, OAuth 2.0/OIDC, Active Directory, or PKI/CAC certificates):

• Identity Mapping & Role-Based Access (RBAC): The MCP server validates the identity of both the prompt submitter and the LLM agent against ICAM policy databases.

• Granular Tool Authorization: The LLM isn't given direct database or system access. Instead, the MCP server exposes specific "tools" (get_camera_feed, fetch_personnel_log) mapped directly to the user's ICAM privilege tier.

• Zero Trust & Security Boundaries: If an unauthorized user prompts the LLM to access physical access logs or trigger a lock-down, the MCP tool checks the ICAM token, rejects the payload at the server layer, and prevents unauthorized actions.

2. Connecting to Equitus Fusion & Arcxa (Data Migration & Integration)

Equitus Fusion (utilizing Knowledge Graph Neural Networks / KGNN) automates the ingestion, ETL, and semantic mapping of disparate data sources into AI-ready graphs. Arcxa provides graph visualization, relationship mapping, and analytics:

• Automated Data Migration & ETL: The LLM uses MCP tools to discover legacy databases, S3 storage buckets, and unstructured files. It automates schema alignment, entity resolution, and data migration into the Equitus unified graph.

• Semantic Graph Queries: Instead of requiring operators to write complex SPARQL/Cypher graph queries, the LLM converts natural language into structured MCP function calls against Equitus Fusion/Arcxa.

• Bidirectional Integration: The MCP layer allows data updates made or summarized by the LLM to write back safely to enterprise applications or downstream systems under explicit human supervision.

3. Connecting to EVS (Equitus Video Sentinel & Physical Security)

EVS processes real-time camera streams for computer vision, object detection, anomaly tracking, and perimeter monitoring:

• Physical Threat Ingestion: EVS generates metadata events ((S) unauthorized vehicle (P) detected at (O)Gate 3, person loitering in restricted server room). The MCP server exposes these visual alerts as real-time context to the LLM.

• Forensic Search & Retrieval: An operator can ask the LLM: "Find all blue trucks that passed through the west perimeter between 2:00 PM and 4:00 PM yesterday." The LLM calls EVS forensic metadata indices via MCP tools to return matching video clips and timeline correlations.

4. Convergence: Cyber & Physical Security Orchestration

By bridging ICAM, EVS, and Arcxa graph intelligence into an LLM via MCP, the system unlocks Cyber-Physical Security Convergence:

• Cross-Domain Threat Correlation:

  • Scenario: EVS detects a physical badge reader swipe at Facility A (Physical Event). Simultaneously, ICAM flags a concurrent Active Directory login for the same user account from an IP address in another country (Cyber Anomaly).

  • LLM Role: The LLM receives correlated telemetry from EVS and ICAM via the MCP server, identifies a credential theft / physical spoofing breach, and alerts security operators with recommended containment actions.

• Automated Incident Response Workflows: Upon detecting a threat, the LLM can execute predefined, supervised MCP write tools—such as revoking user privileges in ICAM, locking down physical access points via EVS/VMS integrations, or opening an incident ticket in the SOC dashboard.

Key Operational Benefits

1. Air-Gapped & On-Premises Readiness: Equitus architectures (RHEL/OpenShift-native) can run on-premise alongside local LLM weights (e.g., Ollama, vLLM, private endpoints), preserving data privacy without cloud risk.

2. Explainability & Provenance: Every response generated by the LLM relies on auditable graph data and EVS visual logs via the MCP server, eliminating hallucinations and ensuring compliance (NERC CIP, CJIS, HIPAA).

3. Controlled Tool Surface: The MCP server acts as a sandbox, preventing LLM prompt injections from executing unvetted operational commands

FUSION MCP - Mechanics

• Subject–Predicate–Object Triples: Every piece of data—whether it is an enterprise employee profile from ICAM, a database schema field during migration, or a bounding box event from EVS video analytics—is broken down into semantic triples ( (S) [Employee: John Doe] ---> (P) [hasBadgeAccess] ---> (O) [Zone: Server Room B]).

• Knowledge Graph Ingestion: Equitus Fusion maps these disparate, siloed data formats into a unified, interconnected graph. Instead of dealing with rigid SQL tables or unstructured JSON logs, the graph represents relationships natively.

• Knowledge Graph Neural Networks (KGNN): KGNNs layer machine learning over the graph topology. They compute node embeddings, detect structural anomalies, predict missing links, and score relationship probabilities across cyber and physical domains.

• The MCP Bridge: The Equitus Fusion MCP Server exposes these graph capabilities, enterprise ICAM rules, and EVS event streams as standardized "tools" and contextual resources to the LLM.

2. MCP  Enables Autonomous Co-Pilots

• Semantic Grounding (Eliminating Hallucinations): Standard LLMs often guess answers based on parametric memory, causing hallucinations. By forcing the LLM to query a semantic triple store via MCP, every assertion the model makes is backed by explicit, verifiable graph edges and node relationships.

• Contextual Data Migration & Integration: When migrating legacy systems, traditional ETL pipelines break when schemas change. A triple store architecture abstracts schemas into universal relationships. The LLM can ask the MCP server to map legacy tables into the unified graph by reasoning about object classes and predicates rather than manual column mapping.

• Cross-Domain Cyber-Physical Correlation: Traditional security keeps IT logs (ICAM/Active Directory) and physical security (EVS video/badges) in separate silos. Because a knowledge graph natively links multi-domain entities (e.g., linking a physical person to an IP address, a badge ID, and a server rack), the LLM can instantly correlate a physical breach with a simultaneous cyber anomaly, executing complex investigations and secure, role-restricted orchestrations in real time.