YugabyteDB MCP Server

What is YugabyteDB MCP Server about?

YugabyteDB MCP Server provides an implementation of the Model Context Protocol (MCP) that connects LLMs with a YugabyteDB instance. It exposes database metadata and read‑only query capabilities through a lightweight server that can be run locally or in a container.

How to use YugabyteDB MCP Server?

1. Clone and install

   git clone https://github.com/yugabyte/yugabytedb-mcp-server.git
   cd yugabytedb-mcp-server
   uv sync
   

2. Configure the connection by setting the YUGABYTEDB_URL environment variable (e.g., in a .env file).
3. Run the server
   • STDIO transport: uv run src/server.py
   • Streamable‑HTTP transport: uv run src/server.py --transport http
4. Add the server to an MCP client (Claude Desktop, Cursor, Windsurf, etc.) using the provided JSON snippets.
5. Interact – ask the client to list tables, summarize schemas, or run SELECT queries; results are returned as JSON.

Key features of YugabyteDB MCP Server

• List all tables with schema definitions and row counts.
• Execute read‑only SQL queries and receive JSON‑formatted results.
• Supports both STDIO and Streamable‑HTTP transports.
• Docker image available for easy deployment.
• Compatibility with FastMCP and major MCP clients (Claude Desktop, Cursor, Windsurf Editor).

Use cases of YugabyteDB MCP Server

• Data exploration: LLMs can quickly summarize database structures for developers.
• Ad‑hoc reporting: Generate JSON results from SELECT statements without writing client code.
• AI‑assisted debugging: LLMs query the DB to verify data integrity during troubleshooting.
• Rapid prototyping: Embed database queries in AI‑driven applications or notebooks.

FAQ from the YugabyteDB MCP Server

• What versions of Python are supported? Python 3.10 or newer.
• Do I need Maven or Node.js? No, the server runs with Python and uv for dependency management.
• Can I run the server in production? Yes, using the Docker image (mcp/yugabytedb) with appropriate environment variables and security hardening.
• How do I enable HTTPS? Use a reverse proxy (e.g., Nginx) in front of the Streamable‑HTTP transport.
• What if uv is not found? Ensure uv is installed and on your PATH; alternatively, symlink it to /usr/local/bin/uv as the troubleshooting guide suggests.
• Which queries are allowed? Only read‑only queries; attempts to modify data will be rejected.

YugabyteDB MCP Server

An MCP server implementation for YugabyteDB that allows LLMs to directly interact with your database.

Features

• List all tables in the database, including schema and row counts
• Run read-only SQL queries and return results as JSON
• Designed for use with FastMCP and compatible with MCP clients like Claude Desktop, Cursor, and Windsurf Editor

Prerequisites

• Python 3.10 or higher
• uv installed to manage and run the server
• A running YugabyteDB database
• An MCP client such as Claude Desktop or Cursor

Installation

Clone this repository and install dependencies:

git clone git@github.com:yugabyte/yugabytedb-mcp-server.git
cd yugabytedb-mcp-server
uv sync

Configuration

The server is configured using the following environment variable:

• YUGABYTEDB_URL: The connection string for your YugabyteDB database (e.g., dbname=database_name host=hostname port=5433 user=username password=password)

Example .env file:

YUGABYTEDB_URL=postgresql://user:password@localhost:5433/yugabyte

Usage

Running the Server

You can run the server with STDIO transport using uv:

uv run src/server.py

or with Streamable-HTTP transport:

uv run src/server.py --transport http

Running the Server with Docker

Build the Docker image:

docker build -t mcp/yugabytedb .

Run the container with STDIO transport:

docker run -p 8080:8080 -e YUGABYTEDB_URL="your-db-url" mcp/yugabytedb

or with Streamable-HTTP transport:

docker run -p 8080:8080 -e YUGABYTEDB_URL="your-db-url" mcp/yugabytedb --transport=http

MCP Client Configuration

To use this server with an MCP client (e.g., Claude Desktop, Cursor), add it to your MCP client configuration.

Running via uv

Example configuration for Cursor:

{
  "mcpServers": {
    "yugabytedb-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/cloned/yugabytedb-mcp-server/",
        "run",
        "src/server.py"
      ],
      "env": {
        "YUGABYTEDB_URL": "dbname=database_name host=hostname port=5433 user=username password=password load_balance=true topology_keys=cloud.region.zone1,cloud.region.zone2"
      }
    }
  }
}

• Replace /path/to/cloned/yugabytedb-mcp-server/ with the path to your cloned repository.
• Set the correct database URL in the env section.

Running via Docker (e.g., in Claude)

After building the docker container, add the following to claude_config.json entry or equivalent json files for other editors:

{
  "mcpServers": {
    "yugabytedb-mcp-docker": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "YUGABYTEDB_URL=dbname=yugabyte host=host.docker.internal port=5433 user=yugabyte password=yugabyte load_balance=false",
        "mcp/yugabytedb"
      ]
    }
  }
}

Claude Desktop

1. Edit the configuration file. Go to Claude -> Settings -> Developer -> Edit Config
2. Add the above configuration under mcpServers.
3. Restart Claude Desktop.

Claude Desktop Logs

The logs for Claude Desktop can be found in the following locations:

• MacOS: ~/Library/Logs/Claude
• Windows: %APPDATA%\Claude\Logs

The logs can be used to diagnose connection issues or other problems with your MCP server configuration. For more details, refer to the official documentation.

Cursor

1. Install Cursor on your machine.
2. Go to Cursor > Settings > Cursor Settings > MCP > Add a new global MCP server.
3. Add the configuration as above.
4. Save the configuration.
5. You will see yugabytedb-mcp-server as an added server in MCP servers list. Refresh to see if server is enabled.

Cursor Logs

In the bottom panel of Cursor, click on "Output" and select "Cursor MCP" from the dropdown menu to view server logs. This can help diagnose connection issues or other problems with your MCP server configuration.

Windsurf Editor

1. Install Windsurf Editor on your machine.
2. Go to Windsurf > Settings > Windsurf Settings > Cascade > Model Context Protocol (MCP) Servers > Add server > Add custom server.
3. Add the configuration as above.
4. Save and refresh.

Streamable-HTTP with MCP Inspector

1. Start the server using Streamable-HTTP:

   uv run src/server.py --transport http
   

   Or with Docker:

   docker run -p 8080:8080 -e YUGABYTEDB_URL="..." mcp/yugabytedb --transport=http
   

2. Launch the inspector:

   npx @modelcontextprotocol/inspector
   

3. In the GUI, use the URL:

   http://localhost:8080/invocations/mcp
   

   • Change transport type to Streamable-HTTP
   • Add the proxy token from the terminal output

Tools Provided

• summarize_database: Lists all tables in the database, including schema and row counts.
• run_read_only_query: Runs a read-only SQL query and returns the results as JSON.

Example Usage

Once connected via an MCP client, you can:

• Ask for a summary of the database tables and schemas
• Run SELECT queries and get results in JSON

Environment Variables

• YUGABYTEDB_URL: (required) The connection string for your YugabyteDB/PostgreSQL database

Troubleshooting

• Ensure the YUGABYTEDB_URL is set and correct
• Verify your database is running and accessible
• Check that your user has the necessary permissions
• Make sure uv is installed and available in your PATH. Note: If claude is unable to access uv, giving the error: spawn uv ENOENT, try symlinking the uv for global access:

sudo ln -s "$(which uv)" /usr/local/bin/uv

• Review logs in your MCP client for connection or query errors

Development

• Project dependencies are managed in pyproject.toml
• Main server logic is in src/server.py