Overview
MCP is an open standard that allows AI applications, such as Infor GenAI, Kiro, Claude, and Cursor, to discover the tools that an MCP Server exposes and invoke them on behalf of users.
With MCP Server support, API Gateway acts as a secure entry point in front of MCP Servers, which are services that implement the Model Context Protocol. These MCP Servers are built by Infor application teams and provisioned through the Registry, delivered through BaaS, or created by customers and third parties. Administrators designate which registered endpoints are MCP Servers so that the gateway can proxy MCP traffic to them, enforce authorization, and make them discoverable to AI agents.
API Gateway does not convert conventional REST or SOAP APIs into MCP Servers. An endpoint designated as an MCP Server must already implement the Model Context Protocol. The designation tells the gateway to route and secure that endpoint as an MCP Server.
What this guide covers
This documentation covers:
- Designating endpoints as MCP Servers
- Creating MCP Client authorized apps for external tools
- Connecting external MCP Clients, including Kiro, Claude, and Cursor
- Using MCP tools in Infor GenAI Agents
- Metadata discovery for MCP-enabled APIs
- Error-handling standards
Who this affects
- IONAPI-Administrator users: Configure MCP Server endpoints and create MCP Client authorized apps.
- GenAI Agent builders with access to Infor GenAI Studio: Add MCP Server tools to agents in Infor GenAI Studio.
- Developers using external MCP Clients: Connect tools such as Kiro, Claude Desktop, or Cursor to Infor and custom MCP Servers through API Gateway.
- API consumers: Discover MCP-enabled endpoints programmatically through API metadata.
Business benefits
MCP Server support enables API Gateway to act as a single, secured gateway in front of the MCP Servers that your organization already runs, whether they are built by Infor application teams, delivered through BaaS, or created by your own teams and third parties. AI agents access these MCP Servers through the gateway instead of connecting directly to each server.
- Single, governed entry point for MCP Servers: Register your MCP Servers in the gateway and expose them to AI agents through one secured endpoint instead of managing direct connections to each MCP Server.
- Centralized, governed access: Access is controlled for each client through the gateway authorization model, enabling administrators to determine which AI clients can access specific MCP Servers.
- Secure user-level authorization: MCP calls run under the authenticated user's identity through OAuth 2.1, so AI agents operate within each user's existing permissions instead of using a shared service account.
- One configuration across multiple clients: The same MCP Servers are accessible to Infor GenAI Agents and external clients, including Kiro, Claude Desktop, and Cursor, reducing duplicated configuration.
Common use cases
- AI-assisted ERP operations: Enable a GenAI agent to answer questions such as “What is the status of order 12345?” by calling an Infor MCP Server, such as M3, through the gateway without requiring a developer to configure a direct connection.
- Cross-system correlation: An agent can combine data from multiple MCP Servers, for example, by correlating delayed orders with warehouse stock levels to answer questions that span applications.
- Developer tooling: Engineers can use external MCP Clients, including Kiro, Cursor, and Claude Desktop, to access Infor and custom MCP Servers through the gateway from their development environments. Access is limited to the servers that they are authorized to use.
- Custom MCP Server exposure: Teams that operate their own MCP Servers can register them as Non-Infor suites and expose them to AI agents through the gateway authorization and discovery model.