Astra DB MCP Server

A Model Context Protocol (MCP) server for interacting with Astra DB. MCP extends the capabilities of Large Language Models (LLMs) by allowing them to interact with external systems as agents.

Prerequisites

You need to have a running Astra DB database. If you don't have one, you can create a free database here. From there, you can get two things you need:

1. An Astra DB Application Token

2. The Astra DB API Endpoint

To learn how to get these, please read the getting started docs.

Related MCP server: MCP Database Server

Adding to an MCP client

Here's how you can add this server to your MCP client.

Claude Desktop

To add this to Claude Desktop, go to Preferences -> Developer -> Edit Config and add this JSON blob to claude_desktop_config.json:

Optional Keyspace Configuration: By default, this server uses the keyspace configured in the underlying Astra DB library (typically default_keyspace). If you need to connect to a specific keyspace, you can add the ASTRA_DB_KEYSPACE variable to the env object above, like so:

Windows PowerShell Users: npx is a batch command so modify the JSON as follows:

Cursor

To add this to Cursor, go to Settings -> Cursor Settings -> MCP

From there, you can add the server by clicking the "+ Add New MCP Server" button, where you should be brought to an mcp.json file.

Tip: there is a ~/.cursor/mcp.json that represents your Global MCP settings, and a project-specific .cursor/mcp.json file that is specific to the project. You probably want to install this MCP server into the project-specific file.

Add the same JSON as indiciated in the Claude Desktop instructions.

Alternatively you may be presented with a wizard, where you can enter the following values (for Unix-based systems):

• Name: Whatever you want

• Type: Command

• Command:

Note:

Once added, your editor will be fully connected to your Astra DB database.

Available Tools

The server provides the following tools for interacting with Astra DB:

Collection Management

• GetCollections: Get all collections in the database

• CreateCollection: Create a new collection in the database (with vector support)

• UpdateCollection: Update an existing collection in the database

• DeleteCollection: Delete a collection from the database

• EstimateDocumentCount: Get estimate of the number of documents in a collection

Record Operations

• ListRecords: List records from a collection in the database

• GetRecord: Get a specific record from a collection by ID

• CreateRecord: Create a new record in a collection

• UpdateRecord: Update an existing record in a collection

• DeleteRecord: Delete a record from a collection

• FindRecord: Find records in a collection by field value

• FindDistinctValues: Find distinct values for a specific field in a collection

Bulk Operations

• BulkCreateRecords: Create multiple records in a collection at once

• BulkUpdateRecords: Update multiple records in a collection at once

• BulkDeleteRecords: Delete multiple records from a collection at once

Vector Search

• VectorSearch: Perform vector similarity search on vector embeddings

• HybridSearch: Combine vector similarity search with text search

Utility

• OpenBrowser: Open a web browser for authentication and setup

• HelpAddToClient: Get assistance with adding Astra DB client to your MCP client

New Features and Capabilities

Vector Search Capabilities

The Astra DB MCP server now includes powerful vector search capabilities for AI applications:

VectorSearch

Perform similarity search on vector embeddings:

HybridSearch

Combine vector similarity search with text search for more accurate results:

Enhanced Collection Creation

The CreateCollection tool now supports more vector configuration options:

Finding Distinct Values

The new FindDistinctValues tool allows you to find unique values for a field:

Optimized Bulk Operations

Bulk operations now use native batch processing for better performance:

Improved Error Handling

The server now provides more detailed error messages with error codes to help diagnose issues more easily.

Changelog

All notable changes to this project will be documented in this file. The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Running evals

The evals package loads an mcp client that then runs the index.ts file, so there is no need to rebuild between tests. You can load environment variables by prefixing the npx command. Full documentation can be found here.

❤️ Contributors

Badges