MCP Resources / Primitives

What Are MCP Resources?

For developers building AI agents, MCP resources define the interfaces between the language model and the external world. Understanding these three primitives — tools, resources, and prompts — is the prerequisite for developing powerful AI integrations. MCP resources form the foundation of the Model Context Protocol ecosystem.

The Model Context Protocol (MCP) defines three fundamental object types — called primitives — through which MCP servers provide functionality. These three primitives form the foundation of every MCP integration and determine how AI applications interact with external data sources and tools.

Tools are executable functions that the LLM can actively call — such as a database query, a calculation, or an API call. They correspond to classic tool calling. Resources provide data and context that the LLM can read — comparable to a file system or database from which the agent retrieves information. Prompts are predefined templates and workflows that the server offers the client as reusable building blocks. Each primitive has its own lifecycle and access controls.

For businesses developing or integrating MCP servers, understanding the three primitives is essential. A well-designed MCP integration uses resources for data access, tools for actions, and prompts for standardized processes. Combined with MCP sampling and RAG systems, this creates a powerful infrastructure that supplies AI agents with all the information and tools they need.