DALLAS, May 05, 2025 (GLOBE NEWSWIRE) — XMPro today announced the release of Multi-Agent Generative System (MAGS) version 1.5, introducing an advanced trust architecture for industrial AI that establishes new standards for reliability, security, and cross-domain collaboration. The update features evidence-based confidence scoring, multi-method consensus decision-making, standardized agent-to-agent communication, and seamless AI integration through Model Context Protocol (MCP).
The new capabilities directly address the critical challenges industrial organizations face when deploying AI systems in environments where reliability, safety, and performance are non-negotiable requirements.
Learn more at http://www.xmpro.com
Watch Introductory Demo here → XMPro’s Collaborative AI Agent Teams For Industrial Operations
Watch the Deep Dive Demo here → Collaborative AI Agent Teams for Autonomous Industrial Operations
Agent-to-Agent (A2A) Communication Protocol
XMPro MAGS v1.5 implements Google’s Agent-to-Agent (A2A) protocol as a communication framework that transforms how AI agents interact across organizational boundaries. A2A establishes a common language for AI agents from different providers to communicate while respecting different trust requirements.
“The A2A protocol integration bridges the traditional divide between operational technology and information technology,” said Gavin Green, VP of Strategic Solutions at XMPro. “Organizations can now maintain high-trust standards for industrial systems while enabling controlled collaboration with business domains that may operate under different requirements.”
XMPro’s implementation uses a layered approach:
- A2A DataStream Connector: Enables no-code configuration of agent communication, maintaining XMPro’s visual approach to agent design
- Protocol Bridge: Translates between XMPro’s existing MQTT/OPC US/DDS/Kafka-based communication and A2A’s JSON-RPC format
- Agent Card Capabilities: Each agent exposes its capabilities through a digital identity that describes what it can do and how to authenticate with it
“What distinguishes industrial AI from general business applications is the need for absolute trust in automated systems that can affect physical operations,” said Pieter van Schalkwyk, CEO at XMPro. “With XMPro MAGS 1.5, we’ve created a comprehensive trust architecture that gives industrial organizations the confidence to deploy AI at scale, while maintaining appropriate boundaries between operational and business domains.”
Evidence-Based Confidence Scoring
MAGS 1.5 introduces a sophisticated confidence assessment framework that evaluates agent observations, reflections, plans, and actions using five key dimensions, including evidence strength, consistency analysis, reasoning quality assessment, uncertainty quantification and stability measurement.
The system combines these factors using configurable weights to produce normalized confidence scores categorized into various confidence levels, allowing organizations to set appropriate thresholds for different types of decisions based on criticality.
Multi-Method Consensus Decision-Making
MAGS 1.5 introduces an advanced consensus framework that enables agent teams to make better decisions together. This system combines:
- Collaborative Iteration: Agents work through structured rounds of proposal and conflict resolution rather than simple voting
- Intelligent Conflict Detection: Automatically identifies resource contentions and interdependencies between agent plans
- Adaptive Protocols: Dynamically selects appropriate decision methods based on situation complexity
- Expertise Weighting: Gives greater influence to agents with relevant domain expertise
- Confidence Integration: Adjusts validation requirements based on confidence scores
- Smart Escalation: Routes low-confidence decisions to humans with comprehensive context
- Complete Traceability: Captures all proposals, conflicts, and justifications for audit purposes
This system reduces decision bottlenecks, improves plan quality, and creates the right balance between agent autonomy and human oversight—enabling teams to tackle complex challenges with greater reliability and transparency.
Model Context Protocol (MCP) Integration
MAGS 1.5 incorporates the Model Context Protocol (MCP) developed by Anthropic as a standardized access layer for AI models to interact with external data sources and tools. In industrial settings, MCP functions as a “translator” that allows AI models to effectively leverage contextual data through three key capabilities including tools, resources and prompts.
XMPro has implemented MCP Action Agents as DataStream connectors, enabling direct integration of MCP-compliant tools within real-time data processing workflows.
Control and Governance at Scale
Underlying these advancements is XMPro’s architectural approach that uses DataStreams as control envelopes for AI agents. This creates a fundamental separation between agent reasoning and action execution, establishing safety boundaries that don’t depend on perfect agent behavior.
“In industrial environments, you can’t rely solely on an agent’s internal constraints,” explained van Schalkwyk. “Our approach allows organizations to deploy sophisticated AI capabilities while maintaining rigorous control over what actions can be executed in their operational environments.”
AI agents in the XMPro MAGS framework can observe data, reflect on patterns, and develop action plans, but they cannot directly execute these actions. Instead, all proposed actions must pass through DataStream control mechanisms that evaluate them against predefined rules and constraints. This separation ensures safety isn’t compromised even if an agent’s reasoning produces inappropriate recommendations.
Strategic Benefits of the Trust Architecture
The comprehensive trust architecture in MAGS 1.5 delivers significant strategic benefits for industrial organizations:
- OT/IT Integration: The longstanding challenge of bridging operational technology and information technology is addressed by enabling a heterogeneous but interoperable ecosystem where each domain maintains its appropriate level of rigor.
- Organizational Coherence Without Compromise: Different parts of the enterprise can work in concert while respecting their distinct trust requirements, eliminating the need to force a single standard across domains with different reliability needs.
- Selective Trust Boundary Control: Industrial organizations can maintain high-trust operational systems while selectively exposing capabilities to business functions through well-defined interfaces.
- Human-AI Collaboration Model: The system identifies when human review is needed, creating an effective collaboration framework where humans remain in control of critical decisions.
- Future-Proofing Across Domains: As the AI agent landscape evolves toward specialization, the standards-based approach positions XMPro MAGS to participate in broader agent ecosystems while maintaining industrial-grade security.
| Decisions & Reasoning Diagram – XMPro Collaborative AI Agent Teams For Autonomous Industrial Operations |
Building on Successful Hannover Messe 2025 Showcase
Earlier this month, XMPro successfully showcased its MAGS framework at Hannover Messe 2025 as part of an integrated demonstration with Dell Technologies. The well-received demonstration highlighted how collaborative AI agent teams can address complex industrial challenges without requiring extensive data science expertise or specialized IT infrastructure.
Availability
MAGS version 1.5 is available immediately for existing customers and will be available to new customers starting May 15, 2025. For more information, visit http://www.xmpro.com or contact sales@xmpro.com.
About
XMPro helps industrial companies rapidly build intelligent operations solutions using composable AI, digital twins, and real-time data streams. Our platform enables collaborative AI agent teams to monitor, reason, and act—turning complex data into actionable intelligence.
Learn more at http://www.xmpro.com/apex-ai
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/8c07f614-9b7d-48ad-9448-7feb3e43696b
