- Saturday, Oct 17th, 2026
| TITLE | Building Production-Scale Agentic AI Systems: A Multi-LLM, MCP, and A2A Architecture for Enterprise Workflow Automation |
|---|---|
| ABSTRACT | Enterprise workflow automation is evolving from conventional rule-based orchestration and isolated generative AI applications toward autonomous systems capable of planning, reasoning, executing, and coordinating complex business activities. Agentic Artificial Intelligence (AI) provides a foundation for this transition by combining large language models (LLMs), specialized AI agents, enterprise tools, and workflow orchestration mechanisms. However, deploying agentic AI at production scale introduces architectural challenges involving model selection, agent coordination, tool interoperability, security, observability, reliability, governance, and cost management. This article presents a generalized production architecture for enterprise agentic AI systems based on a Multi-LLM, Model Context Protocol (MCP), and Agent-to-Agent (A2A) communication model. The proposed architecture separates reasoning, orchestration, tool execution, and inter-agent collaboration into independently governed layers, enabling organizations to combine multiple LLM providers and specialized models according to workload requirements. MCP-based tool connectivity provides a standardized mechanism for agents to interact with enterprise applications, databases, APIs, and operational services, while A2A communication enables specialized agents to collaborate on multi-stage business workflows. The architecture incorporates policy enforcement, identity management, human-in-the-loop controls, observability, fault isolation, model routing, and evaluation mechanisms to support secure and reliable production deployment. The article further examines workflow decomposition, agent lifecycle management, context propagation, failure recovery, scalability, and operational governance. By combining Multi-LLM orchestration with standardized tool interfaces and structured agent collaboration, the proposed approach provides an extensible foundation for automating complex enterprise processes while maintaining operational control, security, and business accountability. |
| AUTHOR | Sreedivya Sharma Staff Software Engineer, Walmart Labs, United States of America |
| VOLUME | 13 |
| DOI | DOI:10.15680/IJARETY.2026.1305006 |
| 6_Building Production-Scale Agentic AI Systems A Multi-LLM, MCP, and A2A Architecture for Enterprise Workflow Automation.pdf | |
| KEYWORDS | |
| References | 1. Anthropic, “Model Context Protocol Specification,†Model Context Protocol, Nov. 25, 2025. 2. Model Context Protocol, “The 2026 MCP Roadmap,†Mar. 9, 2026. 3. Model Context Protocol, “2026-07-28 MCP Specification Release Candidate,†Jul. 28, 2026. 4. Model Context Protocol, “What is the Model Context Protocol?†2026. 5. A2A Project, “Agent2Agent (A2A) Protocol Specification,†Nov. 9, 2025. 6. A2A Project, “Announcing A2A Protocol 1.0,†2025. 7. A2A Project, “What is A2A?†2026. 8. A2A Project, “A2A Protocol Roadmap,†Mar. 10, 2026. 9. National Institute of Standards and Technology, “Announcing the AI Agent Standards Initiative for Interoperable and Secure AI,†Feb. 17, 2026. 10. H. Booth, National Institute of Standards and Technology, “Accelerating the Adoption of Software and AI Agents,†NIST Concept Paper, Feb. 5, 2026. 11. National Institute of Standards and Technology, “Building Evaluation Probes into Agentic AI,†May 1, 2026. 12. National Institute of Standards and Technology, “AI Risk Management Framework,†2026. |
Copyright © IJARETY 2023 All Rights Reserved.