|
Actions
|
MCP Client
|
API Client
|
|
Request Initiation
|
AI Agents act as MCP client to send context-aware requests.
|
Applications use the API client to access services via REST, OData, or HTTP.
|
|
Server Processing
|
All MCP requests are processed through the MCP gateway, where the MCP Server capability is realized. The gateway provides::
- Tools and prompts for AI agents
|
All API requests are processed through the API gateway, which provides:
- Multi-protocol support \(REST, SOAP, and OData\)
- API artifacts for standard APIs
|
|
Backend Integration
|
Requests are routed to:
- SAP LOB cloud
- Third-party APIs \(for example, databases, GitHub, cloud storage\)
- Managed API on the same Integration Cell tenant
|
Requests are routed to:
- SAP LOB cloud
- Third-party APIs \(for example, databases, GitHub, cloud storage\)
|
|
Response Delivery
|
AI agents receive responses in a format that is compliant with the MCP protocol.
|
API clients get standard REST/OData responses.
|
This architecture enables reuse of existing APIs for both AI and API clients—ensuring secure, scalable, and governed access to enterprise systems in both AI-driven and traditional use cases.
## **MCP Server Primitives**
MCP server primitives define the core capabilities that an MCP server exposes to AI agents. They provide standardized ways for agents to perform actions, access contextual data, and use predefined interaction patterns.
An MCP server supports three primary primitives: Tools, Resources, and Prompts.
Each primitive serves a distinct purpose and is exposed through the MCP protocol. Together, they enable structured and controlled interactions between AI agents and enterprise services.
- **MCP Tools**
MCP tools represent actions that an AI agent can invoke to perform specific operations.
They are derived from API operations and can represent actions such as retrieving, creating, updating, or deleting data. Each tool defines an input schema that specifies the parameters accepted from the agent and an output schema that describes the structure of the returned response.
Tools enable AI agents to interact with backend services in a structured and controlled manner.
- **MCP Resources**
MCP resources provide AI agents with access to data and contextual information exposed by an MCP server.
They represent information that agents can retrieve and use as context when performing tasks or generating responses. Resources are identified using URIs and can represent documents, records, configuration data, or other application-specific information.
Unlike tools, resources are intended for retrieving information rather than performing actions.
- **MCP Prompts**
MCP prompts define reusable interaction patterns for AI agents.
They provide predefined messages and instructions that guide how an agent performs a task. Prompts can accept arguments that dynamically customize their content. Messages can use **User** and **Assistant** roles to define a structured interaction flow.
Prompts help ensure consistent and repeatable interactions between AI agents and the MCP server.
**Related Information**
[API Artifact](api-artifact-a4f87b1.md "API artifacts are the core building blocks used to design, configure, publish, and manage APIs in SAP Integration Suite, API Management. These artifacts define how APIs behave, how they interact with backend systems, and how they are exposed to consumers.")
[How Model Context Protocol \(MCP\) Server Enables AI Integration](how-model-context-protocol-mcp-server-enables-ai-integration-572e7fd.md "The MCP server enables AI agents to interact with enterprise systems through a structured process.")
[Create an MCP Server from an API Artifact](create-an-mcp-server-from-an-api-artifact-9f18140.md "Create an MCP Server from an existing API artifact deployed on Integration Cell in SAP Integration Suite. This allows you to expose deployed integration APIs as tools that AI agents and AI clients can discover and invoke.")
[Create an MCP Server from an HTTP Endpoint](create-an-mcp-server-from-an-http-endpoint-b59f141.md "Create an MCP Server from any HTTP endpoint by providing its OpenAPI specification. This enables you to connect and expose REST-based services as MCP tools for AI-driven consumption.")
[Create an MCP Server by Referring to an RFC-Based Backend](create-an-mcp-server-by-referring-to-an-rfc-based-backend-cbb8b78.md "Create an MCP server by connecting to an RFC-based backend system and exposing selected RFC operations as MCP tools.")