Skip to content
/
Tier B A3 MCP-anchored

Microsoft MCP Servers

ms-mcp-servers · microsoft/mcp · ★ 3.2k · last commit 2026-05-26

Primitive shape 40 total
MCP tools 40
00

Summary

microsoft/mcp — Summary

Microsoft MCP Servers is Microsoft's official monorepo for Model Context Protocol server implementations — currently shipping the Azure MCP Server (v1.0 GA) and Microsoft Fabric MCP Server, plus shared engineering infrastructure (test frameworks, engineering systems, pipelines) for Microsoft's MCP contributor ecosystem. The repo also serves as a catalog for all Microsoft-published MCP servers (20+ total, from both this repo and linked external repos).

Problem it solves: AI agents lack native access to Azure cloud services, Microsoft Fabric analytics, and the broader Microsoft product ecosystem. MCP servers expose these services through the standardized MCP protocol so any MCP-compatible AI tool (VS Code Copilot, Claude, IntelliJ, Eclipse) can query and operate Azure/Fabric resources.

Distinctive trait: The Azure MCP Server exposes 40+ Azure services through a single MCP server — the broadest Microsoft-cloud coverage in any MCP server. It is generally available (GA) and production-ready, with enterprise governance (AGENTS.md, CODE_OF_CONDUCT.md, SECURITY.md, NOTICE.txt, Dockerfile, Smithery support). The shared engineering infrastructure in core/ and eng/ reduces duplication across Microsoft's MCP contributor teams.

Target audience: AI developers building Claude/Copilot/etc. integrations that need to query or operate Azure resources, Microsoft Fabric datasets, or other Microsoft services.

Scope: 3,209 GitHub stars, MIT license, 502 forks, C# primary language (dotnet), active maintenance (last push May 26 2026).

Differs from seeds: Unlike any seed framework, microsoft/mcp is an enterprise MCP toolserver (Archetype 3 pattern extended to cloud services). The closest is ccmemory (also Archetype 3 — MCP-anchored) but ccmemory is a memory server (10 tools) while Azure MCP is an Azure operations server (40+ services). No seed covers cloud infrastructure management through MCP.

01

Overview

microsoft/mcp — Overview

Origin

Published by Microsoft (microsoft GitHub organization) in April 2025. The Azure MCP Server reached v1.0 GA status (announced with a blog post). Microsoft chose MCP as their primary AI integration protocol for Azure developer tooling.

Philosophy

The README framing: connecting AI agents to Azure services through the standardized MCP protocol:

"All Azure MCP tools in a single server. The Azure MCP Server implements the MCP specification to create a seamless connection between AI agents and Azure services."

Key principles:

  1. Single-server simplicity: All 40+ Azure services through one MCP server
  2. Multi-IDE support: VS Code, VS Code Insiders, Visual Studio 2026/2022, IntelliJ, Eclipse — all with one-click install
  3. Sovereign cloud support: Handles Azure Government, China cloud variants
  4. Shared engineering infrastructure: core/ and eng/ provide test frameworks, pipelines, and tooling shared across all Microsoft MCP servers — reducing duplication

Catalog vs Source

The README serves two purposes:

  1. Catalog: Lists all Microsoft MCP servers (from this repo + external repos) with VS Code install badges
  2. Source: Azure MCP Server and Fabric MCP Server are developed here

Microsoft MCP Servers Listed (from README, ~20+ total)

Server Type Category
Azure MCP Source (this repo) Cloud & Infrastructure
Microsoft Fabric MCP Source (this repo) Cloud & Infrastructure
Microsoft Foundry MCP Remote (mcp.ai.azure.com) Cloud & Infrastructure
Azure Resource Manager MCP External repo Cloud & Infrastructure
Azure DevOps MCP External repo Developer Tools
Azure Kubernetes Service MCP External repo Cloud & Infrastructure
GitHub MCP External (github/github-mcp-server) Developer Tools
GitHub Awesome-Copilot MCP External Developer Tools
Markitdown MCP External (microsoft/markitdown) Developer Tools
Microsoft 365 Agents Toolkit External Developer Tools
(10+ more in README)
02

Architecture

microsoft/mcp — Architecture

Distribution

  • Type: mcp-server (multiple) + shared engineering infrastructure
  • Language: C# (dotnet) for Azure and Fabric servers
  • License: MIT
  • Install methods: VS Code extension, NuGet, NPM, PyPI, Docker, MCPB (5 package registries)

Repository Structure

microsoft/mcp/
├── core/                          # Shared: C# test framework + MCP libraries
├── servers/
│   ├── Azure.Mcp.Server/          # Azure MCP Server (v1.0 GA)
│   │   ├── src/                   # C# source
│   │   ├── tests/                 # Test suite
│   │   ├── docs/                  # Documentation
│   │   ├── server.json            # MCP server manifest
│   │   ├── README.md              # Server-specific README
│   │   ├── CHANGELOG.md
│   │   ├── TROUBLESHOOTING.md
│   │   └── SUPPORT.md
│   ├── Fabric.Mcp.Server/         # Microsoft Fabric MCP Server
│   └── Template.Mcp.Server/       # Template for new MS MCP servers
├── tools/                         # Shared tooling
├── eng/                           # Engineering systems + CI/CD pipelines
├── docs/                          # Repo-level documentation
├── .config/                       # Configuration
├── .devcontainer/                 # Dev container config
├── .vscode/                       # VS Code workspace settings
├── Directory.Build.props           # MSBuild shared properties
├── Directory.Packages.props        # NuGet centralized package management
├── global.json                    # .NET SDK pinning
├── Dockerfile                     # Container build
├── AGENTS.md                      # AI agent behavior guidelines
├── SECURITY.md
├── NOTICE.txt
├── smithery.yaml                  # Smithery registry config
└── Microsoft.Mcp.slnx             # .NET solution file

Multi-Registry Distribution (Azure MCP)

The Azure MCP Server is available from 5 package registries:

NuGet:   dotnet tool install Azure.Mcp.Server
NPM:     npx @azure/mcp@latest server start
PyPI:    uvx azure-mcp@latest server start
Docker:  docker run mcr.microsoft.com/azure-mcp-server:latest
MCPB:    (mcp tooling protocol)

Plus IDE-specific install:

VS Code: vscode:extension/ms-azuretools.vscode-azure-mcp-server
IntelliJ: plugins.jetbrains.com/plugin/8053
Eclipse: marketplace.eclipse.org/content/azure-toolkit-eclipse

Protocol

  • Transport: stdio (local) + HTTP remote endpoint
  • Remote: https://mcp.azure.com (Azure MCP remote mode)
  • Standard: MCP 2025-12-11 spec

Required Runtime

  • .NET SDK (for C# server)
  • OR docker (for container deployment)
  • OR node/python (for npm/pip install)
  • Azure credentials (for actual Azure operations)
03

Components

microsoft/mcp — Components

Azure MCP Server — Azure Services Covered (40+)

The Azure MCP Server exposes tools for the following Azure service categories:

Category Services
Compute Azure VMs, Container Apps, Azure Functions, AKS
Storage Blob Storage, Table Storage, Queue Storage, File Storage
Databases Cosmos DB, Azure SQL, PostgreSQL, MySQL
AI/ML Azure OpenAI, Azure AI Foundry, Cognitive Services
Networking Virtual Networks, DNS, Traffic Manager, Application Gateway
Security Key Vault, Managed Identity, Defender for Cloud
Monitoring Azure Monitor, Log Analytics, Application Insights
DevOps Azure DevOps Boards, Repos, Pipelines
Infrastructure Resource Groups, Resource Graph, ARM Templates
Serverless Logic Apps, Event Grid, Event Hubs, Service Bus
Data Azure Data Factory, Synapse Analytics, Data Lake
Identity Entra ID (Azure AD), RBAC

Microsoft Fabric MCP Server

A separate MCP server for Microsoft Fabric (the unified analytics platform). Focuses on:

  • Fabric workspace operations
  • Dataset and report management
  • Data lakehouse access

Template MCP Server (Template.Mcp.Server)

A reference template for Microsoft teams building new MCP servers — enforces shared patterns, test structure, and deployment configuration.

Shared Engineering Infrastructure

core/

  • C# base classes for MCP server development
  • Test framework (shared across Azure and Fabric servers)
  • Common MCP protocol utilities

eng/

  • CI/CD pipeline definitions
  • Release engineering scripts
  • Build automation

tools/

  • Developer productivity tools for the repo

AGENTS.md

Microsoft ships an AGENTS.md file in the repo root — providing AI-coding-agent guidelines for contributors working in this codebase. This is governance documentation for AI agents contributing to MCP servers.

05

Prompts

microsoft/mcp — Prompts

Note on Prompt Content

The Microsoft MCP servers are tool infrastructure — C# server implementations that expose Azure APIs through the MCP protocol. They do not ship prompt files or skill instructions. The "prompting" happens at the AI client layer when users describe what they want to do with Azure.

Verbatim Excerpt 1: AGENTS.md — AI Agent Guidelines for Contributors

File: AGENTS.md Technique: Coding agent governance document

[Content not extracted — file exists but content requires gh API access]

Purpose: AGENTS.md provides behavioral guidelines for AI coding agents (like Claude Code or GitHub Copilot) when they contribute to the microsoft/mcp codebase. This is a meta-prompt — instructions to the AI about how to contribute code.

Verbatim Excerpt 2: Azure MCP README — Tool Description Pattern

File: servers/Azure.Mcp.Server/README.md

# Azure MCP Server

All Azure MCP tools in a single server. The Azure MCP Server implements the MCP
specification to create a seamless connection between AI agents and Azure services.
Azure MCP Server can be used alone or with the GitHub Copilot for Azure extension
in VS Code.

## What can you do with the Azure MCP Server?

Supercharges your agents with Azure context across **40+ different Azure services**.

Pattern: The README is itself the primary "prompt" surface — it describes capabilities in natural language that AI agents can read to understand what the server does and how to invoke it.

Verbatim Excerpt 3: server.json — MCP Manifest

File: servers/Azure.Mcp.Server/server.json

{
  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
  "name": "Azure MCP Server",
  "description": "All Azure MCP tools to create a seamless connection between AI agents and Azure services.",
  "version": "...",
  "packages": [
    { "registryType": "nuget", "identifier": "...", "transport": { "type": "stdio" }, ... },
    { "registryType": "npm", "identifier": "...", "transport": { "type": "stdio" }, ... },
    { "registryType": "pypi", "identifier": "...", "transport": { "type": "stdio" }, ... }
  ]
}

Pattern: The MCP server manifest is a structured declaration following the 2025-12-11 MCP spec schema — not a prompt, but the machine-readable discovery mechanism that MCP clients use to understand and install the server.

09

Uniqueness

microsoft/mcp — Uniqueness & Positioning

Differs from Seeds

Unlike all 11 seeds (which are coding agent frameworks), microsoft/mcp is an enterprise MCP toolserver extending AI agents with Azure cloud operations. The closest seed is ccmemory (Archetype 3 — MCP-anchored) but ccmemory is a memory management tool (~10 tools) while Azure MCP is a cloud operations platform (40+ services). No seed covers cloud infrastructure through MCP.

The Azure MCP Server most resembles claude-flow's MCP component (305 tools) in the breadth of capability exposed through a single MCP server — but Microsoft's scope is narrower (Azure specifically, not general agent orchestration).

Distinctive Properties

  1. Broadest single-service MCP coverage: 40+ Azure services through one server — the widest cloud coverage of any MCP server in production
  2. Five-registry distribution: NuGet + NPM + PyPI + Docker + MCPB — the only framework in the corpus distributed through 5 different package registries simultaneously
  3. Six-IDE install badges: VS Code, VS Code Insiders, Visual Studio 2026, Visual Studio 2022, IntelliJ, Eclipse — broadest IDE support in the batch
  4. Shared engineering infrastructure: core/ + eng/ + Template.Mcp.Server enable Microsoft's internal MCP teams to build consistently — a multiplier for Microsoft's MCP ecosystem
  5. AGENTS.md governance: Ships AI-coding-agent governance guidelines for contributors — rare for a vendor MCP repo
  6. Sovereign cloud support: Azure Government and Azure China support — enterprise compliance requirement not addressed by any OSS framework

Observable Failure Modes

  • C# language barrier: Primary contributors need .NET/C# skills — higher barrier than Python/TypeScript for most AI developers
  • Azure credentials dependency: Every tool call requires valid Azure auth — cannot be tested without an Azure account
  • Broad scope = broad attack surface: 40+ Azure services means 40+ potential misconfiguration vectors
  • Version lag: With frequent Azure API changes, tool implementations may lag behind new Azure features

Microsoft MCP Ecosystem

The repo serves as the hub for Microsoft's broader MCP strategy — cataloging 20+ Microsoft MCP servers from across the organization. This catalog function is as valuable as the server implementations, as it provides a single discovery point for all Microsoft MCP tools.

04

Workflow

microsoft/mcp — Workflow

User Workflow (Agent Consumer)

The MCP server enables a query-and-operate workflow for AI agents:

1. INSTALL MCP SERVER
   VS Code: one-click from vscode:extension/ms-azuretools.vscode-azure-mcp-server
   OR: npx @azure/mcp@latest server start
   OR: docker run mcr.microsoft.com/azure-mcp-server

2. CONFIGURE AI CLIENT
   Add to mcp.json or settings.json:
   { "servers": { "azure": { "type": "stdio", "command": "npx", "args": ["@azure/mcp@latest", "server", "start"] } } }

3. AUTHENTICATE
   Azure CLI: az login
   OR: Managed Identity (in cloud)
   OR: Service Principal credentials

4. USE IN AGENT
   "List all my Azure resource groups"
   "Show me the blob containers in storage account X"
   "Get the scaling settings for my Container App"
   "Query Log Analytics workspace for errors in the last hour"

No Development Lifecycle

This is a tool server, not a workflow framework. There is no phase sequence, no approval gates, no spec documents. The MCP server responds to tool invocations from the agent.

Release Workflow (for contributors)

1. Create feature branch
2. Implement in C# following shared core patterns
3. Add tests following shared test framework
4. Submit PR → CI/CD runs in eng/ pipelines
5. Release through 5 package registries (NuGet, NPM, PyPI, Docker, MCPB)
6. Update CHANGELOG.md with changelog-entries/

Sovereign Cloud Support

The Azure MCP Server explicitly handles Azure sovereign clouds:

  • Azure Government
  • Azure China (operated by 21Vianet)
  • Connection strings and endpoints differ; the server handles routing

Artifacts

Phase Artifact
Install MCP server process (stdio or HTTP)
Query Tool response JSON (MCP standard)
Release Versioned packages across 5 registries
06

Memory Context

microsoft/mcp — Memory & Context

No Agent-Side Memory

The Azure MCP Server is stateless — it does not maintain memory of past interactions. Each tool call is independent.

State Lives in Azure

The "state" accessed through Azure MCP is in Azure itself:

  • Resource configurations in Azure Resource Manager
  • Log Analytics workspace data
  • Key Vault secrets
  • Blob storage contents

The MCP server reads/writes this state through Azure APIs but does not cache or persist it locally.

Azure Credentials (Authentication State)

The server requires authenticated Azure credentials:

  • Azure CLI (az login) — local credential cache
  • Managed Identity (cloud) — no credential files
  • Service Principal — environment variables or config files

The credentials are the persistent state required for operation.

Context for AI Clients

The server returns structured JSON responses to MCP tool calls. AI clients (VS Code Copilot, Claude, etc.) handle their own context management. The server does not dictate how clients use its responses.

Sovereign Cloud Context

The server handles multiple Azure environments through endpoint configuration — the context of which cloud is active is a configuration concern, not session memory.

07

Orchestration

microsoft/mcp — Orchestration

Single-Agent Target

The Azure MCP Server is designed for single-agent use — one AI assistant calling Azure tools. There is no multi-agent coordination at the MCP server level.

Orchestration Pattern

None — the server responds to individual tool invocations. Orchestration (if any) happens at the AI client layer.

Isolation Mechanism

Process (when run as stdio server) or container (when run via Docker). Each server instance is isolated from other processes.

Multi-Model

No — the MCP server is model-agnostic. It exposes tools that any MCP-compatible AI client can use, regardless of the underlying model.

Execution Mode

Event-driven — the server responds to incoming tool call requests. No background processing, no scheduled operations.

Protocol

  • Local (stdio): Server runs as a child process of the AI client
  • Remote (HTTP): Server exposed at https://mcp.azure.com for remote access

Cross-Client Portability

High — works with any MCP-compliant client:

  • VS Code (GitHub Copilot)
  • Claude (via Claude Code MCP config)
  • IntelliJ IDEA
  • Eclipse IDE
  • Visual Studio 2022/2026
  • Any HTTP MCP client
08

Ui Cli Surface

microsoft/mcp — UI/CLI Surface

Dedicated CLI Binary

No standalone CLI binary — the Azure MCP Server is a server process, not a user CLI. It is invoked by the MCP client (VS Code, Claude Code, etc.), not by the user directly.

Server start command (used by MCP clients):

npx @azure/mcp@latest server start

IDE Integration (Multi-IDE)

IDE Integration Method Install
VS Code Extension (ms-azuretools.vscode-azure-mcp-server) One-click
VS Code Insiders Same extension One-click
Visual Studio 2026 GitHub Copilot for Azure extension One-click
Visual Studio 2022 Same extension (marketplace) One-click
IntelliJ IDEA Azure Toolkit plugin One-click
Eclipse IDE Azure Toolkit for Eclipse One-click

All have VS Code-style install badges in the README ([![Install Azure MCP in VS Code](...)](vscode:extension/...)).

Remote Endpoint

https://mcp.azure.com

Any MCP-compliant client can connect:

{ "type": "http", "url": "https://mcp.azure.com" }

Observability

The Azure MCP Server itself does not ship an observability dashboard. Observability comes from:

  • Azure Monitor (for operations performed through the server)
  • VS Code Output panel (MCP server logs)
  • Standard .NET logging

Dev Container

.devcontainer/ provides a VS Code dev container for contributors — GitHub Codespaces-compatible setup for developing MCP servers in a consistent environment.

Smithery Support

smithery.yaml registers the server with the Smithery MCP server registry — an alternative discovery mechanism to GitHub.

Related frameworks

same archetype · same primary tool · same memory type

Taskmaster AI ★ 27k
A3 MCP-anchored

Converts a PRD into a dependency-ordered JSON task graph that AI coding agents execute one task at a time, eliminating context…

ccmemory ★ 1
A3 MCP-anchored

Accumulates decisions, corrections, and failed approaches from Claude Code sessions into a queryable Neo4j graph so each new…

Pimzino spec-workflow-mcp ★ 4.2k
A3 MCP-anchored

MCP server providing spec-driven development workflow with dashboard-backed approval gates, implementation logging, and VSCode…

MCP Shrimp Task Manager ★ 2.1k
A3 MCP-anchored

Convert natural language requests into structured AI development tasks with chain-of-thought enforcement, reflection gates, and…

Bernstein ★ 460
A3 MCP-anchored

Govern parallel CLI coding agents with a deterministic Python scheduler, HMAC-chained audit trail, and compliance-ready signed…

LeanSpec ★ 252
A3 MCP-anchored

Provides a unified spec CLI and MCP server over any existing spec backend (markdown, GitHub Issues, ADO), making spec-driven…