Jun 30, 2025
Model Context Protocol (MCP)
- Paola Di Maio1
- 1CSKRNS
- Human Cell Atlas Method Development Community
- Bio Neuromod
Protocol Citation: Paola Di Maio 2025. Model Context Protocol (MCP) . protocols.io https://dx.doi.org/10.17504/protocols.io.3byl46ebzgo5/v1
License: This is an open access protocol distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Protocol status: Working
We use this protocol and it's working
Created: June 30, 2025
Last Modified: June 30, 2025
Protocol Integer ID: 221289
Keywords: AI, agents, protocols, model context protocol, mcp, protocol, standardized open protocol, managing context, secure interaction, context across session, secure interaction between ai model, model
Abstract
The Model Context Protocol (MCP) is a standardized open protocol designed to enable
seamless, secure interaction between AI models and external tools, resources, and systems by
managing context across sessions.
Guidelines
Review and test on a case by case basis
Troubleshooting
Server Creation Phase ○ Establish MCP servers that expose Tools (functions callable by AI models), Resources (data endpoints), and Prompts (templates guiding tool usage). ○ Register these components with standardized metadata to enable discovery and invocation by clients. ○ Set up secure communication channels and authentication mechanisms to protect resources. (This phase initializes the MCP server lifecycle and prepares it for operation) .
Client-Server Communication and Session Establishment ● MCP clients (embedded within AI applications or hosts) initiate connections to MCP servers using a defined client-server architecture. ● Clients negotiate capabilities and maintain stateful sessions to preserve context across multiple interactions. ● Communication uses structured JSON-RPC messages over secure transport protocols, ensuring integrity and confidentiality .
Authentication and Authorization ● Implement user authentication to verify human users (e.g., via OAuth 2.1 with PKCE). ● Authorize AI clients to access specific tools/resources on behalf of users, enforcing fine-grained access control. ● Use well-defined OAuth endpoints (/authorize, /token, and discovery URLs) to manage secure token exchange and session validation .
Tool Invocation and Context Management ● AI models invoke registered tools via JSON-RPC calls, passing relevant context and parameters. ● MCP servers execute the requested tools and return structured results or error messages. ● Context is managed incrementally, allowing persistent memory and continuity across agent actions, addressing the “disconnected models” problem in AI systems .
Server Operation and Update Phase ● MCP servers continuously operate, handling multiple client requests, managing sessions, and monitoring security. ● Servers support dynamic updates, including adding/removing tools, updating prompts, and refining security policies without disrupting ongoing sessions. ● Security and privacy risks are actively mitigated through monitoring and protocol enhancements .
Ecosystem Integration and Extension ● MCP supports integration with diverse AI models, multi-agent systems, and external platforms, enabling interoperability and scalability. ● Extensions include support for streaming responses, multi-modal data, and advanced coordination among agents
Protocol references
Hou, X., Zhao, Y., Wang, S., & Wang, H. (2025). Model Context Protocol (MCP):
Landscape, Security Threats, and Future Research Directions. arXiv preprint
arXiv:2503.23278. https://doi.org/10.48550/arXiv.2503.23278
Krishnan, N. (2025). Advancing Multi-Agent Systems Through Model Context Protocol:
Architecture, Implementation, and Applications. arXiv preprint arXiv:2504.21030.
Philschmid. (2025, April 3). Model Context Protocol (MCP) an overview.
●oecc, A. (2025, May 20). Model Context Protocol: Landscape, Security Threats, and
Future. LinkedIn.
-yoecc