Overview of CloudHealth Model Context Protocol (MCP)
search cancel

Overview of CloudHealth Model Context Protocol (MCP)

book

Article ID: 457554

calendar_today

Updated On:

Products

CloudHealth

Issue/Introduction

As AI integrations become more prevalent, users need a reliable way for external AI systems to interact with CloudHealth.
Traditional APIs (like GraphQL) are designed for human developers and lack the structured contracts required for AI to operate deterministically.
This article explains the CloudHealth Model Context Protocol (MCP), why it is necessary, and how it differs from native AI offerings like Intelligent Assist.

Resolution

The CloudHealth Model Context Protocol (MCP) is an AI-friendly API abstraction layer that exposes governed, structured tools to AI systems.

Why MCP instead of GraphQL?

While GraphQL is highly effective for standard development, AI systems require explicit execution rules and structured tool contracts. Standardizing on an MCP layer prevents customers from needing to build their own AI abstraction, authentication, and orchestration layers. Additionally, MCP provides:

  • Precise logging and auditing of AI-originated calls.

  • AI-specific rate limiting and governance controls (distinguishing AI traffic from human or script traffic).

  • A clear AI interaction boundary.

MCP vs. Intelligent Assist

  • Intelligent Assist is CloudHealth’s native AI interface.

  • MCP enables customers to use their own private AI systems (such as Claude Desktop), connect custom AI agents through a governed integration layer, and control where model processing occurs.

Dataset Selection and RAG Integration

To prevent hallucinations, tool descriptions and system guidance instruct the AI to call getDatasets before constructing a query, ensuring it uses valid, tenant-specific fields.

When visualizing how MCP works alongside Retrieval-Augmented Generation (RAG) and the AI model itself, use the following mental model:

  • RAG = Memory (Provides product documentation and knowledge)

  • MCP = Hands (Executes structured data and actions)

  • Orchestrator / AI Model = Brain (Acts as the reasoning layer)