mcp-neon-azure-ai-agent

# Azure AI Agent Service, MCP and Neon Integration This project connects **Azure AI Agent Service** with **Neon Serverless Postgres** using **MCP (Model Context Protocol)**. It lets you send commands to an AI agent that can create a Neon management Agent to manage Neon projects, databases, and more.  ## What It Does - MCP server integrates with Azure AI Foundry to enable connections to your existing Azure AI Agents. - MCP tool creates a new Neon management agent in the Azure AI Foundry. Starts an AI Agent with tools like `create_project` (via Neon API). - Lets you ask the agent to do tasks like creating a Neon project. - Uses MCP to communicate with the agent in real time. ## Why Use This? - Use natural language to control your Neon database - Automate project and database management - Build powerful RAG (Retrieval-Augmented Generation) apps using Neon ## How It Works 1. When the server starts, it: - Loads environment variables - Connects to Azure AI Agent Service - Creates an AI Agent with tool functions (like `create_project`) 2. The agent listens for queries through MCP 3. When a query is sent, it: - Creates a thread - Adds a message - Runs the tool automatically - Returns the result ## Requirements - Python 3.9+ - An **[Azure AI Foundry](https://learn.microsoft.com/en-us/azure/ai-services/agents/quickstart?pivots=ai-foundry-portal)** setup - A **[Neon API Key](https://neon.tech/docs/manage/api-keys#creating-api-keys?refcode=44WD03UH)** - MCP installed and running. Refer to the [Claude Desktop Users get started guide](https://modelcontextprotocol.io/quickstart/user) ## Setup 1. Clone this repository: ```bash git clone https://github.com/neondatabase-labs/mcp-neon-azure-ai-agent.git cd mcp-neon-azure-ai-agent ``` 2. Create a new virtual environment: ```bash python -m venv venv && source venv/bin/activate ``` 3. Install dependencies: ```bash pip install -r requirements.txt ``` 4. Set up environment variables by creating a `.env` file: ```ini PROJECT_CONNECTION_STRING=your_azure_project_connection_string AZURE_OPENAI_CHAT_DEPLOYMENT_NAME=your_deployment_name NEON_API_KEY=your_neon_api_key ``` 5. Configure Claude Desktop (or another MCP client) Update your MCP client configuration (claude_desktop_config.json) to integrate the MCP server with the following in any text editor: ```json { "mcpServers": { "azure-agent": { "command": "/ABSOLUTE/PATH/TO/neon-mcp-azure-ai-agent/venv/bin/python", "args": [ "-m", "azure_agent_mcp_server" ], "cwd": "/ABSOLUTE/PATH/TO/neon-mcp-azure-ai-agent", "env": { "PYTHONPATH": "/Users/boburumurzokov/Neon/neon-mcp-azure-ai-agent", "PROJECT_CONNECTION_STRING": "eastus.api.azureml.ms;7a478544-9d2b-4f43-9916-7787f55aa58d;ai-agents-demo;mcp-neon-demo", "AZURE_OPENAI_CHAT_DEPLOYMENT_NAME": "gpt4o", "NEON_API_KEY": "napi_oigyl7kyhd69jmi3w3tlmfs6zbrna2u3hi1d3s94clputmvmoa1c6ej30p1536al" } } } } ``` Replace `/ABSOLUTE/PATH/TO` to your valid directory path where `neon-mcp-azure-ai-agent` folder is located. 6. Restart Claude After updating your configuration file, you need to restart Claude for Desktop. Upon restarting, you should see a hammer icon in the bottom right corner of the input box. After clicking on the hammer icon, you should see the tools. ## Try a Query From your MCP client or Claude App interface, try: ```text Create a Neon database project named 'My MCP'. ``` ## What's Next - Add more Neon tools (create_branch, create_database, etc.) - Log results - Support SQL queries --- Built with ❤️ using Azure + Neon + MCP ## Other useful resources - [Neon MCP Server](https://neon.tech/docs/ai/neon-mcp-server) - [Build AI Agents with Azure AI Agent Service and Neon](https://neon.tech/blog/build-ai-agents-with-azure-ai-agent-service-and-neon)