Skip to main content
Glama

Astra MCP Server

by smatiolids
GitHub
Python
MIT License
1
OverviewInspectNewEndpointsSchemaRelated ServersReviewsScore
Need Help?View Source CodeReport Issue

Astra MCP Server

A Model Context Protocol (MCP) server for interacting with Astra DB (DataStax Astra).

Astra MCP Server Overview

Running it as MCP Server with STDIO

To run the astra-mcp-server as MCP Server with STDIO, you can use the following command:

uvx astra-mcp-server -tr stdio --astra_token <astra_token> --astra_endpoint <astra_endpoint>

Running it as MCP Server with HTTP

To run the astra-mcp-server as MCP Server with STDIO, you can use the following command:

uvx astra-mcp-server

App options

  • --transport <transport>: The transport to use for the MCP Server. Valid values are stdio and http. Default is stdio.

  • --astra_token <astra_token>: The Astra token to use for the Astra DB connection. If not filled, the app will try to get the token from the ASTRA_DB_APPLICATION_TOKEN environment variable.

  • --astra_endpoint <astra_endpoint>: The Astra endpoint to use for the Astra DB connection. If not filled, the app will try to get the endpoint from the ASTRA_DB_API_ENDPOINT environment variable.

  • --tags <tags>: The tags to filter the tools for the MCP Server.

Catalog options

  • --catalog_file <catalog_file>: The catalog file to use for the MCP Server. Default is tools_config.json. If not filled, the app will try to get the catalog from the ASTRA_DB_CATALOG_COLLECTION environment variable.

  • --catalog_collection <catalog_collection>: The catalog collection to use for the MCP Server. Default is tool_catalog. If not filled, the app will try to get the catalog from the ASTRA_DB_CATALOG_COLLECTION environment variable.

Options valid for the HTTP transport:

When running the app as HTTP, you can use the following options:

  • --host <host>: The host to use for the MCP Server. Default is 127.0.0.1.

  • --port <port>: The port to use for the MCP Server. Default is 8000.

  • --workers <workers>: The number of worker processes to use for the MCP Server. Default is 1.

  • --log-level <log_level>: The log level to use for the MCP Server. Valid values are debug, info, warning, and error. Default is info.

  • --log-file <log_file>: The log file to use for the MCP Server. Default is logs/astra_mcp_server.log.

1. Set up environment variables

If you are running the app with some MCP Client that allows you to define environment variables, you can set the same variables in the MCP Client.

If you prefer, you can create a .env file in app directory with the following variables:

# Astra DB Configuration ASTRA_DB_APPLICATION_TOKEN=your_astra_db_token ASTRA_DB_API_ENDPOINT=your_astra_db_endpoint ASTRA_DB_CATALOG_COLLECTION=tool_catalog LOG_LEVEL=DEBUG LOG_FILE=./logs/logs.log # Logging Configuration (optional) LOG_LEVEL=INFO LOG_FILE=logs/astra_mcp_server.log #Embedding Configuration OPENAI_API_KEY=your_openai_api_key OPENAI_BASE_URL=your_openai_base_url IBM_WATSONX_API_KEY=your_ibm_watsonx_api_key IBM_WATSONX_BASE_URL=your_ibm_watsonx_base_url IBM_WATSONX_PROJECT_ID=your_ibm_watsonx_project_id

Creating tools

Tools are created based on a json specification. It can be save to a file or to a Astra DB collection (preferable for production use cases).

A tools specifiction json needs the following fields:

{ "tags": ["products"], // The tags of the tool - Filter if needed "type": "tool", // The type of the tool "name": "search_products", // The name of the tool "description": "Search for products", // The description of the tool to the MCP Client "limit": 10, // The limit of the tool "method": "find_documents", // The method to use to execute the tool "collection_name": "products", // The collection to use to execute the tool "projection": {"$vectorize": 1, "metadata": 0}, // The projection of the tool "parameters": [ // The parameters of the tool { "param": "search_query", // The name of the parameter "description": "Query to search for products", // The description of the parameter "attribute": "$vectorize", // The attribute of the parameter... or $vectorize "type": "string", // The type of the parameter "required": 1, // Whether the parameter is required "operator": "$eq", // The operator to use to filter the parameter - if not filled, the operator is $eq "enum": ["baggage", "boarding", "check-in", "flight-status", "other"], // The enum of the parameter "embedding_model": "text-embedding-3-small" // The embedding model to use to generate the embedding }, { "param": "in_stock", // The name of the parameter "value": true, // The value of the parameter - IF FILLED, THE PARAMETER IS NOT SENT TO THE MCP CLIENT and applied by the server "attribute": "in_stock" // The attribute of the parameter } ], }

Save the json document to the file or to the Astra DB collection. When the server is started, it will load the tools from the file or the Astra DB collection.

After storing on Astra DB, the tools will appear in the Astra DB collection like this:

MCP Tool stored on Astra

Updating tools

To update the tools, you can update the json document and save it to the file or to the Astra DB collection.

python catalog.py -f tools_config_example.json -t tool_catalog

Local Development

# Install dependencies uv sync # Run the server uv run astra-mcp-server --host 127.0.0.1 --port 5150 --reload --log-level debug

Alternative: Direct uvicorn (for development)

# Run directly with uvicorn uv run uvicorn server:main --factory --reload --port 5150 --log-level debug

Using MCP Inspector (STDIO)

npx @modelcontextprotocol/inspector uv run astra-mcp-server --log-level debug -tr stdio

This server cannot be installed

-
security - not tested
A
license - permissive license
-
quality - not tested

How are these scores calculated?

remote-capable server

The server can be hosted and run remotely because it primarily relies on remote services or has no dependency on the local environment.

Enables interaction with DataStax Astra DB through the Model Context Protocol. Provides database connectivity and operations for Astra DB instances via secure token-based authentication.

  1. Running it as MCP Server with STDIO
    1. Running it as MCP Server with HTTP
      1. App options
        1. Catalog options
          1. 1. Set up environment variables
            1. Creating tools
              1. Updating tools
                1. Local Development
                  1. Alternative: Direct uvicorn (for development)
                  2. Using MCP Inspector (STDIO)

                New MCP Servers

                MCP directory API

                We provide all the information about MCP servers via our MCP API.

                curl -X GET 'https://glama.ai/api/mcp/v1/servers/smatiolids/astra-mcp-server'

                If you have feedback or need assistance with the MCP directory API, please join our Discord server