home / mcp / kevodb mcp server

KevoDB MCP Server

An MCP server for accessing your data on Kevo

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "kevodb-kevo-mcp": {
      "url": "http://localhost:9000/mcp",
      "headers": {
        "KEVO_HOST": "localhost",
        "KEVO_PORT": "50051"
      }
    }
  }
}

You can expose KevoDB operations through a dedicated MCP server so AI agents and tools can interact with KevoDB using a standardized, UTF-8 based API. This server supports core KevoDB actions like key-value operations, scans, transactions, batches, and statistics, all accessible via MCP tools and a simple runtime configuration.

How to use

Start by running the KevoDB server if it isn’t already running. Then launch the MCP server so AI agents can connect and perform operations like get, put, scan, and transactions against KevoDB. Your agents connect to the MCP endpoint and issue the available tools. The MCP server also exposes a default HTTP endpoint where clients can connect to issue commands, and a local runtime path you can run directly for development and testing.

How to install

Prerequisites you need before installing the MCP server are Python 3.8 or newer and a running KevoDB instance. You will also need the FastMCP library and the Python Kevo SDK to enable MCP functionality.

Install dependencies with the following command.

pip install fastmcp python-kevo

Install KevoDB MCP server dependencies and run

Ensure KevoDB is running on the default endpoint or configure the host/port as needed. You can run the MCP server directly from its entry script.

Start the MCP server with the following command.

python main.py

Configure connections and environment

You can connect KevoDB to the MCP server using environment variables that define where KevoDB is reachable. By default, the MCP server expects KevoDB on localhost port 50051, but you can override this to suit your network.

KevoDB connection defaults can be overridden with these variables.

KEVO_HOST=localhost
KEVO_PORT=50051

Using with AI agents

AI agents that support MCP can connect to the MCP endpoint and use the exposed tools to interact with KevoDB. The server provides a comprehensive set of tools for everyday database operations as well as transactional workflows.

Notes and troubleshooting

If you need to adjust where the MCP server listens, update KEVO_HOST and KEVO_PORT to point to the desired KevoDB instance. If the KevoDB server changes its network location, ensure the MCP server is restarted so the new endpoint is used.

Compatibility and capabilities

These MCP tools provide full coverage of core KevoDB features, including basic key-value operations (get, put, delete), scans (range, prefix, suffix), transactions, batch writes, and statistics retrieval.

Available tools

connect

Establish a connection to the MCP server from your AI agent so subsequent tool calls can be executed.

get

Retrieve a value by key from KevoDB.

put

Store a key-value pair in KevoDB.

delete

Delete a key from KevoDB.

scan

Scan keys with range, prefix, or suffix options.

batch_write

Execute multiple operations in a single batch.

get_stats

Retrieve database statistics.

begin_transaction

Begin a new transaction and obtain a transaction ID.

commit_transaction

Commit a transaction by its ID.

rollback_transaction

Rollback a transaction by its ID.

tx_put

Put a key-value pair within an active transaction.

tx_get

Get a value by key within a transaction.

tx_delete

Delete a key within a transaction.

cleanup

Close the KevoDB connection and clean up resources.