Creativity · MCP — server

MCP Cassandra Server

The MCP Cassandra Server bridges Apache Cassandra's distributed wide-column store with MCP clients. It surfaces keyspaces, column families, and CQL query execution as tools an LLM can call — letting Claude help debug schema issues, write CQL, and inspect row counts across a cluster.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python
Transports: stdio

Capabilities

Tools: list_keyspaces, list_tables, describe_table, execute_cql
Supports authenticated connections and SSL
Connects to self-hosted Cassandra or DataStax Astra

Install

pip install mcp-cassandra

Configuration

{
  "mcpServers": {
    "cassandra": {
      "command": "uvx",
      "args": ["mcp-cassandra"],
      "env": {
        "CASSANDRA_CONTACT_POINTS": "127.0.0.1",
        "CASSANDRA_USERNAME": "cassandra",
        "CASSANDRA_PASSWORD": "<password>",
        "CASSANDRA_KEYSPACE": "analytics"
      }
    }
  }
}

Frequently asked questions

What does the MCP Cassandra server do?

It exposes CQL queries and schema introspection from a Cassandra cluster to MCP clients, so an LLM can explore keyspaces and help write CQL.

Does it support DataStax Astra?

Yes — configure the secure connect bundle path and Astra token in the environment; the driver handles TLS and authentication.

Are there safety rails?

Use a read-only role and avoid granting DROP or TRUNCATE. Cassandra's lack of transactions makes destructive operations hard to reverse.

Sources

Apache Cassandra Documentation — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20