Architectural Principles for Multi-Agent Systems Based on The Model Context Protocol
Abstract
The article examines the architectural role of the Model Context Protocol in the design of multi-agent systems intended to support robust coordination between cognitive and infrastructural components. The relevance of the study is determined by the rapid transition from isolated LLM interactions to distributed agentic environments, in which direct coupling between agents and external APIs, data sources, and tools increases system interdependence, complicates maintenance, and weakens governability. The purpose of the article is to provide a conceptual substantiation of the architectural principles of multi-agent systems based on MCP and to identify their significance for scalability, interoperability, security, and maintainability. The scientific novelty of the study lies in interpreting MCP not merely as an auxiliary integration protocol, but as an autonomous architectural layer that establishes a formalized topology of interaction among hosts, clients, and servers. The principal findings demonstrate that the use of MCP ensures a strict separation of concerns, declarative context management, modular composability, and governance-by-design in capability access. The article concludes that such an approach enhances observability, reduces integration fragility, and creates the foundation for the evolutionary development of multi-agent solutions while preserving control and auditability. The article will be useful for researchers, AI system architects, developers of agentic platforms, and specialists in enterprise digital infrastructure.
Keywords
Downloads
References
- Anthropic. (2024). Introducing the Model Context Protocol. Anthropic. https://www.anthropic.com/news/model-context-protocol
- Anthropic. (2025a). Architecture Overview - Model Context Protocol. https://modelcontextprotocol.io/docs/learn/architecture
- Anthropic. (2025b). Introduction - Model Context Protocol. Anthropic. https://modelcontextprotocol.io/
- Anthropic. (2025c). Security best practices - Model context protocol. Anthropic. https://modelcontextprotocol.io/docs/tutorials/security/security_best_practices
- Arslan, M., Ghanem, H., Munawar, S., & Cruz, C. (2024). A Survey on RAG with LLMs. 28th International Conference on Knowledge Based and Intelligent Information and Engineering Systems (KES 2024), 246, 3781–3790. https://doi.org/10.1016/j.procs.2024.09.178
- Ferrag, M. A., Tihanyi, N., Hamouda, D., Maglaras, L., Lakas, A., & Debbah, M. (2025). From prompt injections to protocol exploits: Threats in LLM-powered AI agents workflows. ICT Express, 12(2), 353–383. https://doi.org/10.1016/j.icte.2025.12.001
- Gokkaya, B., Aniello, L., & Halak, B. (2025). Software supply chain: A taxonomy of attacks, mitigations and risk assessment strategies. Journal of Information Security and Applications, 97, 104324. https://doi.org/10.1016/j.jisa.2025.104324
- Guo, T., Chen, X., Wang, Y., Chang, R., Pei, S., Chawla, N. V., Wiest, O., & Zhang, X. (2024). Large Language Model Based Multi-agents: A Survey of Progress and Challenges. Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence Survey Track, 8048–8057. https://doi.org/10.24963/ijcai.2024/890
- Li, M. Q., & Fung, B. C. M. (2025). Security concerns for Large Language Models: A survey. Journal of Information Security and Applications, 95, 104284. https://doi.org/10.1016/j.jisa.2025.104284
- Qu, C., Dai, S., Wei, X., Cai, H., Wang, S., Yin, D., Xu, J., & Wen, J.-R. (2024). Tool Learning with Large Language Models: A Survey. Frontiers of Computer Science, 19, 198343. https://doi.org/10.1007/s11704-024-40678-2
- Rosic, M., Sumic, D., & Males, L. (2025). Semantic Interoperability of Multi-Agent Systems in Autonomous Maritime Domains. Electronics, 14(13), 2630. https://doi.org/10.3390/electronics14132630
- Staub, N., Haki, K., Aier, S., & Winter, R. (2022). Governance Mechanisms in Digital Platform Ecosystems: Addressing the Generativity-Control Tension. Communications of the Association for Information Systems, 51(1). https://doi.org/10.17705/1CAIS.05137
- Xu, W., Huang, C., Gao, S., & Shang, S. (2025). LLM-Based Agents for Tool Learning: A Survey. Data Science and Engineering, 10, 533–563. https://doi.org/10.1007/s41019-025-00296-9
- Zhan, Q., Liang, Z., Ying, Z., & Kang, D. (2024). InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents. Findings of the Association for Computational Linguistics: ACL 2022, 10471–10506. https://doi.org/10.18653/v1/2024.findings-acl.624