🌟 Microsoft Learn MCP Server
Stop AI Hallucinations. Give your AI assistant (Claude, Cursor, Copilot, Codex, ...) direct access to the latest official Microsoft documentation.
✨ Free. One-click install. No key needed.
🎯 Why install this?
Stop relying on outdated training data or risky web searches. Learn MCP server provides secure, direct access to Microsoft official docs.
-
🧠 Eliminate Hallucinations. Stop your AI from inventing non-existent Azure SDK methods or hallucinating library packages. Get code that actually compiles.
-
🔌 Plug & Play (No Auth). No API keys, no logins, no sign-ups required. Just one-click install and start coding immediately.
-
🛡️ 100% Trusted & Safe. Protect your supply chain. Unlike generic web searches that may scrape insecure blogs or malicious sites, this tool only accesses official 1st-party Microsoft documentation.
-
💸 Completely Free. High search capacity tailored for seamless, heavy coding sessions.
✨ Example Prompts
"Give me the Azure CLI commands to create an Azure Container App with a managed identity."
"Is gpt-5.4 available in Azure EU regions?"
"Are you sure this is the right way to implement
IHttpClientFactoryin a .NET 8 minimal API?"
"Show me runnable Python code to do harms eval using the Azure AI Foundry evaluation SDK."
🌐 The Microsoft Learn MCP Server Endpoint
The Microsoft Learn MCP Server is accessible to any IDE, agent, or tool that supports the Model Context Protocol (MCP). Any compatible client can connect to the following remote MCP endpoint:
https://learn.microsoft.com/api/mcp
Note: This URL is intended for use within a compliant MCP client via Streamable HTTP, such as the recommended clients listed in our Getting Started section. It does not support direct access from a web browser and may return a
405 Method Not Allowederror if accessed manually. For developers who need to build their own solution, please follow the mandatory guidelines in the Building a Custom Client section to ensure your implementation is resilient and supported.
Standard config works in most clients:
{
"servers": {
"microsoft-learn": {
"type": "http",
"url": "https://learn.microsoft.com/api/mcp"
}
}
}
For experimental features, see the Experimental Features section below.
🧪 Experimental Features
The Microsoft Learn MCP Server offers experimental features that are under active development. These features may change or be refined based on user feedback and usage patterns.
OpenAI-Compatible Endpoint
For applications that require OpenAI Deep Research model compatibility, you can use the OpenAI-compatible endpoint:
https://learn.microsoft.com/api/mcp/openai-compatible
This endpoint supports OpenAI Deep Research models and follows the OpenAI MCP specification.
Token Budget Control
To manage token usage and control costs, you can append the maxTokenBudget query parameter to the MCP endpoint URL. This parameter limits the token count in search tool responses by truncating the content to meet your specified budget.
https://learn.microsoft.com/api/mcp?maxTokenBudget=2000
Note: These experimental features are subject to change. We welcome feedback through our GitHub Discussions.
🛠️ Currently Supported Tools
💻 Microsoft Learn CLI
The @microsoft/learn-cli package gives you terminal access to the same tools — search docs, fetch pages, and find code samples — without an MCP client.
# Run instantly (no install)
npx @microsoft/learn-cli search "azure functions timeout"
# Or install globally
npm install -g @microsoft/learn-cli
# then use `mslearn`
mslearn search "azure functions timeout"
Installing the npm package alone does not install agent discovery. The CLI can install a CLI-first skill for the same ecosystems supported by this repository's plugins: GitHub Copilot, Claude Code, and Codex.
# User profile (default)
mslearn setup --cli
# Current repository
mslearn setup --cli --project
# Override auto-detection with explicit targets
mslearn setup --cli --copilot
mslearn setup --cli --claude
mslearn setup --cli --codex
mslearn setup --cli --copilot --project
mslearn setup --cli --claude --project
mslearn setup --cli --codex --project
# Multiple agents
mslearn setup --cli --copilot --claude --codex
Remove only managed discovery content with:
mslearn remove --cli
mslearn remove --cli --copilot
mslearn remove --cli --claude --codex --project
When no target is specified, setup detects installed agents from their well-known user or project
directories. Removal detects only agents with Microsoft Learn CLI-managed discovery artifacts.
Explicit targets (--copilot, --claude, or --codex) override detection and can be combined. This
workflow does not configure MCP or install agents outside the plugin ecosystems such as Cursor.
When both integrations are installed, agents should prefer the Microsoft Learn MCP tools and use
the standalone CLI skill only as a fallback.
Pass --json to get structured JSON output, useful for programmatic processing:
mslearn search "azure openai" --json | jq '.results[].title'
See cli/README.md for the full command reference.
🤖 Agent Skills
Agent Skills are portable instruction packages that help AI agents use tools more effectively. We provide three skills that guide agents on when and how to use the Microsoft Learn MCP tools:
Quick Setup
Choose the integration that matches how you want the agent to retrieve Microsoft Learn content:
The repository plugin also supports Claude Code. Run the following command and restart Claude Code:
/plugin install microsoft-docs@claude-plugins-official
For the MCP-first GitHub Copilot CLI plugin, run:
/plugin install microsoftdocs/mcp
Otherwise:
- Install the MCP Server first — See Installation below
- Copy the skill folders to your project's
.github/skills/or.claude/skills/directory:microsoft-docs— for concepts, tutorials, and factual lookupsmicrosoft-code-reference— for API lookups, code samples, and troubleshootingmicrosoft-skill-creator— meta-skill for generating custom skills about Microsoft technologies
Supported Agents
Agent Skills work across multiple AI agents:
- VS Code (Insiders) — enable
chat.useAgentSkillssetting - GitHub Copilot CLI & Copilot coding agent
- Claude Code, Cursor, OpenAI Codex, and more
Which Skill Do I Need?
🔌 Installation & Getting Started
The Microsoft Learn MCP Server supports quick installation across multiple development environments. Choose your preferred client below for streamlined setup:
⚠️ Building a Custom Client
If your use case requires a direct, programmatic integration, it is essential to understand that MCP is a dynamic protocol, not a static API. The available tools and their schemas will evolve.
To build a resilient client that will not break as the service is updated, you should adhere to the following principles:
- Discover Tools Dynamically: Your client should fetch current tool definitions from the server at runtime (e.g., using
tools/list). Do not hard-code tool names or parameters.- Refresh on Failure: Your client should handle errors during
tool/invokecalls. If a tool call fails with an error indicating it is missing or its schema has changed (e.g., an HTTP 404 or 400 error), your client should assume its cache is stale and automatically trigger a refresh by callingtools/list.- Handle Live Updates: Your client should listen for server notifications (e.g.,
listChanged) and refresh its tool cache accordingly.
❓ Troubleshooting
💻 System Prompt
Even tool-friendly models like Claude Sonnet 4 sometimes fail to call MCP tools by default; use system prompts to encourage usage.
Here's an example of a Cursor rule (a system prompt) that will cause the LLM to utilize microsoft-learn more frequently:
## Querying Microsoft Documentation
You have access to MCP tools called `microsoft_docs_search`, `microsoft_docs_fetch`, and `microsoft_code_sample_search` - these tools allow you to search through and fetch Microsoft's latest official documentation and code samples, and that information might be more detailed or newer than what's in your training data set.
When handling questions around how to work with native Microsoft technologies, such as C#, F#, ASP.NET Core, Microsoft.Extensions, NuGet, Entity Framework, the `dotnet` runtime - please use these tools for research purposes when dealing with specific / narrowly defined questions that may occur.