azure_mcp_server

Azure MCP Server exposes a suite of tools to interact with Azure Blob Storage, Cosmos DB (NoSQL API), and App Configuration through the Model Context Protocol. You can perform common data plane operations such as creating, listing, reading, updating, and deleting containers, blobs, and Cosmos DB items, as well as read and write configuration values. All actions are logged to the audit://azure-operations endpoint for traceability. To use the server, install and run it in your preferred environment (e.g., via Smithery or manual UV-based execution) and issue MCP tool calls through Claude Desktop or any MCP client that adheres to the protocol. The available tools are organized by service: Blob Storage operations like blob_container_create, blob_upload, blob_read, and blob_list; Cosmos DB operations including container and item create/read/update/delete/query; and App Configuration operations such as app_configuration_kv_read, app_configuration_kv_write, and app_configuration_kv_delete. Each operation requires specific parameters (e.g., container_name, blob_name, item, partition_key) and the Cosmos DB notes emphasize the need for a correctly defined partition key and case-sensitive identifiers.

Prerequisites:\n- Python 3.8+ (or a compatible Python environment)\n- Access to the repository for azure_mcp_server (via Git)\n- Optional: Azure CLI for initial authentication if you prefer shell-based login\n\nInstallation steps:\n1) Clone the repository:\nbash\ngit clone <repository-url>\ncd azure-mcp-server\n\n\n2) Prepare your environment:\nIf you use a virtual environment (recommended):\nbash\npython -m venv venv\nsource venv/bin/activate # macOS/Linux\nvenv\n\n# On Windows use: venv\Scripts\activate\n\n\n3) Install dependencies (adjust as needed if you have a requirements.txt or pyproject.toml):\nbash\npip install -r requirements.txt # if present\n# or install with: pip install -e . depending on project setup\n\n\n4) Run the MCP server via UV (as per README):\nbash\nuv --directory /path/to/repo/azure-mcp-server run azure-mcp-server\n\n\n5) Configure credentials: set the environment variables described in the mcp_config (AZURE_STORAGE_ACCOUNT_URL, AZURE_COSMOSDB_ENDPOINT, AZURE_COSMOSDB_KEY, AZURE_APP_CONFIGURATION_ENDPOINT) or authenticate with Azure CLI if preferred. You can also export them in a .env file read by your shell or your runtime environment.\n\n6) Verify installation: ensure the server starts and available MCP endpoints respond to test tool calls.\n

Tips and common issues:\n- Ensure your partition keys for Cosmos DB operations are correctly defined and that the partition_key value you supply matches the item being read/updated/deleted.\n- Cosmos DB and Blob Storage endpoints are case-sensitive and must be provided exactly as configured in Azure.\n- If environment variables are not provided, the server can leverage DefaultAzureCredential which may fall back to Azure CLI credentials after az login.\n- For security, treat the Cosmos DB key as a sensitive secret and avoid committing it to version control. Consider using a secret manager or environment variable injection in your deployment environment.\n- When using Claude Desktop, ensure your claude_desktop_config.json points to the correct directory and server name (azure-mcp-server) as shown in the example.\n- If you encounter MCP debugging issues, consult the official MCP debugging documentation for troubleshooting steps and example calls.\n