Microsoft just announced an MCP Server for Microsoft Learn. This MCP server makes accurate information from Microsoft’s official documentation available for search and retrieve from any AI tool that supports MCP.
According to the documentation on GitHub these are the key capabilities of the MCP server:
- High-Quality Content Retrieval: Search and retrieve relevant content from Microsoft Learn, Azure documentation, Microsoft 365 documentation, and other official Microsoft sources.
- Semantic Understanding: Uses advanced vector search to find the most contextually relevant documentation for any query.
- Optimized Chunking: Returns up to 10 high-quality content chunks (each max 500 tokens), with article titles, URLs, and self-contained content excerpts.
- Real-time Updates: Access the latest Microsoft documentation as it’s published.
Connect to the MCP server from Microsoft Copilot Studio
- In your Copilot Studio agent click “add tool” on the overview page to add a connection to the MCP server
2. Click the “New Tool” button and then click “Custom Connector” on the next screen
In Copilot Studio a custom Connector is used to connect to a MCP server. Clicking the “Model Context Protocol” choice will only bring you to a Microsoft Lear page describing the process of adding MCP as a tool
3. Clicking “Custom Connector” brings you to a new page for adding a Custom Connector
4. Before creating the connector we will need to create a YAML file with information about how to connect to the Microsoft Learn MCP server endpoint.
5. Use an editor like Visual Studio Code (or Notepad) and save the following information as “learnmcp.yaml”
swagger: '2.0'
info:
title: Microsoft Learn
description: Tilgang til Microsoft Learn dokumentasjon
version: 1.0.0
host: learn.microsoft.com
basePath: /
schemes:
- https
paths:
/api/mcp:
post:
summary: Microsoft Learn MCP API
x-ms-agentic-protocol: mcp-streamable-1.0
operationId: InvokeMCP
responses:
'200':
description: Success6. Click “New custom connector” and select “Import an OpenAPI file”
7. Give your connector a name, “Microsoft Learn MCP” and point to the YAML file you created in previous steps
8. Click the “Continue” button and the click “Create Connector” on the next screen
9. Congratulations, you have created a custom connector for the Microsoft Learn MCP server. Time to add the connector as a tool in Copilot Studio.
10. Go back to the Copilot Studio tab in your browser and hit the “Refresh” button
11. Look for your newly created connector under “Model Context Protocol”. Sometimes it takes a while for it to show up. You might have to hit refresh in your browser and try again.
12. When it shows up, select it
13. This brings you to the connect screen where you will have to create a new connection for the connector
14. Click “Create” on the connect to screen
15. Click “add to agent” to finish adding the Microsoft Learn MCP server as a tool in your agent
16. The Microsoft Learn MCP server should now show up in the Tools section in your agent and is ready for use
Let’s test the Microsoft Learn MCP server in Copilot Studio
Navigate to the test pane of our agent and run an example query like “What are the best practices for Power Platform environments according to official Microsoft Learn documentation”
The first time you run a query you will have to connect to the custom connector you created
After creating the connection, hit the Retry button and your query should run. The agent will use the connection to the Microsoft Learn MCP server to get a qualified answer.
At the moment the Microsoft Learn MCP server have one tool available “Microsoft_docs_search” that performs a semantic search against Microsoft official technical documentation.
For more information visit Microsoft’s GitHub documentation for the MCP server at GitHub – MicrosoftDocs/mcp
