Azure-Samples/mcp

**Build AI agents that call real APIs — fast.** Whether you're a C#, Python, Java, or JavaScript developer, this hub helps you build, run, or reuse Model Context Protocol (MCP) servers on Azure to power your AI agents.
---
🚀 1. Run Your Own MCP Server
Host your own MCP server using Azure Functions — pick your language:
- **C#**
- **Python**
- **TypeScript**
---
🧠 2. Build Agents That Use MCP
Use official SDKs or your favorite AI frameworks to write agents that connect to any MCP server.
✅ Official MCP SDKs (Build Clients, Servers, or Tools in Any Language)
- **C# SDK**
- **Python SDK**
- **TypeScript SDK**
- **Java SDK**
🤖 AI SDK Integrations
- **Python: OpenAI Agents SDK** — Use MCPServerSse with Azure OpenAI
- **C#: Semantic Kernel** — Add MCP tools to your Kernel instance
- **JavaScript: LangChain.js** — Use MCP tools in LangChain or LangGraph agents
- **Java: Spring AI** — Plug into Spring Boot with auto-config
- **Azure AI Agents** — Add Bing + Azure AI Search to your agent
---
🧩 3. Use Plug-and-Play MCP Servers
Don't want to build your own tools? These servers expose real APIs via MCP so your agents can just call them.
> ⚠️ These servers are open source and maintained by the community. Microsoft does not review, support, or endorse them. Use at your own discretion.
🗃️ Data APIs
- **Redis** — Access Azure Redis
- **PostgreSQL** — Query Azure PostgreSQL
- **MySQL** — Use with Azure MySQL
- **MongoDB** — Talk to Cosmos DB with MongoDB
- **Azure Data Explorer (ADX)** — Run KQL queries via agent
🔧 Dev & Infra APIs
- **Azure CLI** — Run CLI commands with natural language
- **Kubernetes** — Control AKS clusters
- **GitHub** — Access GitHub APIs via chat or agent
- **Azure DevOps (Python)** — Work item support
- **Azure DevOps (Node.js)** — Repos, pipelines, and boards
---
📚 Learn More
- 📄 Model Context Protocol Spec