Azure Updates MCP Server

An MCP (Model Context Protocol) server that provides AI assistants with seamless access to Azure service updates, retirements, and feature announcements.

Japanese version: README.ja.md

Features

• 🔍 Better Keyword Matching - Full-text search (FTS5) across titles and descriptions with BM25 relevance ranking, offering improved query search than the original API.

• 🎯 Simplified Filtering - Filter by tags, categories, products, dates, and status—no OData knowledge needed.

• ⚡ Fast Response - Local SQLite replication ensures <500ms query response times.

• 🔄 Automatic Sync - Differential synchronization keeps data fresh (configurable interval, default 24h).

• 📊 Help Resource - Expose all available filters and data freshness to help AI construct queries.

Related MCP server: Hermes Search MCP Server

Legal Notice

⚠️ Important: This server accesses the Azure Updates API, which is subject to Microsoft APIs Terms of Use. By using this software, you agree to:

• Comply with all applicable Microsoft API terms and conditions

• Not use the data for advertising or marketing purposes

• Implement appropriate data protection measures

• Respect Microsoft's intellectual property rights

This software is provided under the MIT License for the code implementation only. The data accessed through the Azure Updates API remains subject to Microsoft's terms.

Additional Resources:

• API Usage Policies: docs/azure-updates-api-manual.md

• Privacy Statement: docs/PRIVACY.md

Setup

Option 1: Global Installation (Recommended)

Install globally from GitHub:

npm install -g github:juyamagu/azure-updates-mcp-server

Then configure VS Code (.vscode/mcp.json):

{
  "servers": {
    "azure-updates": {
      "command": "azure-updates-mcp-server"
    }
  }
}

⚠️ Note: On the first run, data synchronization to the local cache will occur, which may take a few minutes before search queries can be served. Once synchronization is complete, subsequent queries will respond quickly.

Option 2: Using npx

For quick testing without installation:

{
  "servers": {
    "azure-updates": {
      "command": "npx",
      "args": ["-y", "github:juyamagu/azure-updates-mcp-server"]
      // Optional environment variables:
      // "env": {
      //   "DATABASE_PATH": "${workspaceFolder}/.azure-updates/data.db",
      //   "SYNC_STALENESS_HOURS": "24",
      //   "LOG_LEVEL": "info"
      // }
    }
  }
}

Note: npx caches packages which may consume more disk space over time. Global installation is recommended for regular use.

For Windows Users

If npm install ... fails on Windows (due to path length limitations or tsc errors), try manual installation from the .tgz file.

Copy the URL of the latest .tgz file from the Releases page and install directly. This eliminates the need for local builds.

# Replace the URL with the actual latest release
npm install -g https://github.com/juyamagu/azure-updates-mcp-server/releases/download/v2.1.0/azure-updates-mcp-server-2.1.0.tgz

Usage

All you need is to ask your AI assistant to use the search_azure_updates tool with natural language queries and optional filters.

What are the latest security updates related to OAuth authentication in Azure since January 2025? Up to 10 results, please.

Your assistant automatically constructs the appropriate parameters and calls the MCP server like this:

{
  "tool": "search_azure_updates",
  "parameters": {
    "query": "OAuth",
    "filters": {
      "tags": ["Security"],
      "modifiedFrom": "2025-01-01"
    },
    "limit": 10
  }
}

Sample agent definitions for GitHub Copilot can be found in:

• .github/agents/azupdates.search.agent.md

• .github/agents/azupdates.ask.agent.md

Configuration

Configuration is done via environment variables. Create a .env file or set environment variables:

See .env.example for all configuration options.

Available Tools

search_azure_updates

Search and filter Azure updates to find relevant items. Returns metadata only (no descriptions) for token efficiency.

Example Parameters:

{
  "query": "OAuth",
  "filters": {
    "tags": ["Security"],
    "modifiedFrom": "2025-01-01"
  },
  "limit": 10
}

get_azure_update

Retrieve complete update details including full Markdown description and URL.

Example Parameters:

{
  "id": "536699"
}

Recommended Workflow:

1. Use search_azure_updates to find relevant updates

2. Use get_azure_update to fetch full details for selected items

Architecture

sequenceDiagram
    participant LLM as AI Assistant
    participant MCP as MCP Server
    participant Search as Search Service
    participant DB as SQLite DB
    participant Sync as Sync Service
    participant API as Azure API

    Note over MCP,DB: Startup
    MCP->>DB: Initialize database
    MCP->>Sync: Check data staleness
    alt Data is stale (>24h)
        Sync->>API: Fetch updates (differential)
        API-->>Sync: Return new/modified updates
        Sync->>DB: Store updates
    end

    Note over LLM,DB: Query Execution
    LLM->>MCP: search_azure_updates<br/>(natural language query)
    MCP->>Search: Execute search
    Search->>DB: FTS5 query + filters
    DB-->>Search: Matching results
    Search-->>MCP: Ranked results
    MCP-->>LLM: JSON response (<500ms)

    Note over LLM,MCP: Metadata Access
    LLM->>MCP: Get azure-updates://guide
    MCP-->>LLM: Available filters & metadata

Local SQLite replication with FTS5 full-text search provides fast queries (<500ms). Differential sync keeps data fresh from Azure Updates API.

Technology Stack:

• MCP SDK: @modelcontextprotocol/sdk

• Database: better-sqlite3 with FTS5

• Testing: Vitest with TypeScript strict mode

• Runtime: Node.js 18+

Development

See Development Guide for setup, testing, and contributing instructions.

Troubleshooting

See Troubleshooting Guide for common issues and solutions.

Documentation

• Installation from Tarball - Package-based installation

• Troubleshooting - Common issues

• Development Guide - Contributing and testing

• Azure Updates API Manual - API reference

License

MIT License - see LICENSE file for details.