Model Context Protocol
An open protocol that standardizes how AI applications expose and consume tools, resources, prompts, and contextual capabilities across client-server boundaries.
Language-Model Context
Model Context Protocol defines a client-server contract for exposing tools, resources, prompts, and other contextual capabilities to AI applications. Its protocol model uses JSON-RPC together with capability negotiation and lifecycle rules so integrations do not need a unique wire contract for every tool provider.
Security Boundary
MCP does not make a tool safe or authorize an action by itself. User consent, authentication, authorization, least privilege, input validation, and policy enforcement remain responsibilities of the surrounding system.
Related LLM Concepts
- Tool Calling
- Resource
- Prompt Injection
- JSON-RPC
Source
- https://modelcontextprotocol.io/specification/2025-11-25
Related technical article: Artificial Intelligence: Philosophy, Theory and Practice.