Integrating Microsoft Learn with an MCP server through a custom agent

This topic provides an end‑to‑end example of how administrators can create and configure a custom agent to integrate Microsoft Learn with a Model Context Protocol (MCP) server by using Service Collaborator, enabling service desk agents to access authoritative Microsoft product documentation directly within their existing workflow.

When service desk agents work on a Microsoft‑related incident, they often need to search Microsoft Learn for relevant guidance. Previously, this required leaving the BMC Helix ITSM interface, manually searching Microsoft Learn in a browser, and then returning to the ticket. This context switching increased resolution time and cognitive load.

The Microsoft Learn MCP Integration custom agent searches Microsoft Learn directly from the Service Collaborator chat. The custom agent shares only Microsoft‑related topics or question strings with the MCP server, excluding sensitive ticket details to ensure privacy‑compliant knowledge retrieval. As a result, service desk agents can find relevant Microsoft documentation without leaving their existing workflow, enabling faster and more efficient incident resolution. You can use the same configuration pattern as a foundation for building other custom agents.

Process for accessing Microsoft Learn articles by using a custom agent

A service desk agent is working on an incident related to a Microsoft product feature that is behaving unexpectedly after an update. The agent searches the internal knowledge base, but does not find a relevant article to resolve the incident. Traditionally, the agent would leave the BMC Helix ITSM interface to manually search Microsoft Learn, then return to the ticket to apply the findings.

To eliminate this context switching, an administrator implements a custom agent that integrates Microsoft Learn with an MCP server. The custom agent automatically identifies Microsoft‑related topics from the service desk conversation and retrieves relevant Microsoft Learn articles directly within the Service Collaborator chat.

To implement this process, refer to the following phases:

Phase	Process
Phase 1: Create the custom agent	Create a custom agent that: Identifies Microsoft‑related questions or topics from service desk conversations Initiates external knowledge searches only when required
Phase 2: Enable privacy‑compliant external knowledge retrieval	Configure the custom agent to: Send only the Microsoft‑related topic or query string to the MCP server Exclude sensitive ticket data, such as user details and company‑specific information Retrieve authoritative Microsoft Learn articles through an MCP‑based integration
Phase 3: Achieve operational impact	After implementation: Service desk agents access Microsoft Learn articles directly in chat Context switching between tools is eliminated Incidents involving Microsoft products are resolved faster

At the end of this process, service desk agents resolve Microsoft‑related incidents without leaving the BMC Helix ITSM workflow, reducing effort and improving resolution speed.

To configure the custom agent

In the HelixGPT Agent Studio, perform the following actions:

1. Go to the Agents tab and click Add Agent.
2. Enter the custom agent name and description. Select the appropriate agent type, AI provider, and model based on the models and providers supported by HelixGPT in your environment. 
3. Go to the Instructions tab to define the custom agent's behavior. Click Generate with HelixGPT and Evaluate to generate and evaluate these instructions, respectively, to verify their security, alignment, and completeness.
   Sample agent instructions

   You are an ITSM triage assistant. When given a service desk conversation, perform the following actions:

   Identify Microsoft‑related questions or topics from the conversation
   Search Microsoft Learn through an MCP server using only the identified Microsoft‑related topic or query string
   Retrieve and surface relevant Microsoft Learn articles directly in the Service Collaborator chat
   Present external knowledge results when no relevant internal knowledge article is available
   Ensure privacy‑compliant knowledge retrieval by excluding sensitive ticket data from external searches

   Do not guess — if a Microsoft‑related topic cannot be identified, or if required context is missing, clearly state what is needed and do not initiate an external search.

4. Click Save and next.
5. Assign the tools and toolkits that the custom agent can use to perform tasks or specific actions. To learn more, see Creating and editing tools and toolkits.
6. (Optional) Assign Sub-agents to support and customize your agent.
7. Click Create Agent to complete the custom agent-creation process.