Skip to main content
The Upsun Cloud Model Context Protocol (MCP) Server allows users to perform infrastructure operations, monitor environments, and manage resources using natural language commands.You can set up the Upsun Cloud MCP Server with this step-by-step guide.
Model Context Protocol (MCP) is a standard interface that allows large language models (LLMs) to communicate with external tools and data sources. With MCP, developers and tool providers integrate once and ensure interoperability across any MCP-compatible system. For comprehensive information about the Model Context Protocol, including specifications, examples, and implementation details, we recommend you visit the official MCP documentation maintained by Anthropic.

​
Get started with MCP on Upsun Cloud

​
Why connect LLMs to external systems?

LLMs typically lack direct access to real-time or system-specific data. To provide current, relevant responses—such as live configuration settings or API data—LLMs need to connect with external systems. Every service offers a different API, schema, and authentication. Without a standard, managing multiple integrations becomes complex, brittle, and error-prone.

​
Standardizing LLM interactions with MCP

MCP standardizes how LLMs interact with external services. Developers create a single MCP integration point that can interface with any MCP-compatible provider. Similarly, tool and data providers only need to expose their service via an MCP interface once—making it accessible to any MCP-enabled application. Think of MCP as the USB-C of LLM integration—one interface that works uniformly across devices.

​
MCP Hosts, Clients, and Servers

MCP operates on a client–server model:
  • MCP Host: The AI application environment (e.g., your IDE, ChatGPT-like interface, or AI agent).
  • MCP Client: The connection created by the host to talk to a service.
  • MCP Server: The external service that exposes an MCP interface.
To connect to multiple external services, a single host needs to manage multiple MCP clients, each connecting to a different MCP server.

​
Additional resources

  • Deep Dive into MCP’s Interaction Model
    Explore MCP’s three core interaction types—prompts (user-driven), resources (application-driven), and tools (model-driven)—to design more powerful and flexible AI workflows.

​
Summary table

Last modified on October 6, 2026