SingleStore MCP Server

What is SingleStore MCP Server about?

Provides an MCP server that acts as a bridge between Large Language Model (LLM) clients (Claude, Cursor, VS Code Copilot, etc.) and SingleStore's management services. Users can issue natural‑language commands to retrieve information, control workspaces, execute SQL queries, and manage notebooks/jobs.

How to use SingleStore MCP Server?

1. Install – the package is published on PyPI. Install via uvx (recommended) or pip:

   uvx singlestore-mcp-server start   # runs the server directly
   # or
   pip install singlestore-mcp-server
   

2. Configure an MCP client – add the following JSON to the client’s mcpServers configuration:

   {
     "mcpServers": {
       "singlestore-mcp-server": {
         "command": "uvx",
         "args": ["singlestore-mcp-server", "start"]
       }
     }
   }
   

   The server will launch a browser‑based OAuth flow for authentication; no API keys or environment variables are required for the standard setup.
3. Docker alternative – when running inside Docker you must supply an API key via MCP_API_KEY.
4. Interact – invoke the /help prompt in your MCP‑enabled client to see the list of available tools.

Key Features

• Retrieve current user and organization details.
• Interactive organization selection (when not using an API key).
• List and manage workspace groups and individual workspaces.
• Resume, create, and terminate starter workspaces.
• Execute arbitrary SQL queries on any workspace.
• Create, upload, and manage notebook files in SingleStore Spaces.
• Define scheduled jobs from notebooks and manage job lifecycle.
• Fully compatible with major MCP clients (Claude Desktop/Code, Cursor, VS Code Copilot, Windsurf, Gemini CLI, LM Studio, Goose, Qodo Gen).

Use Cases

• Data Exploration – ask an LLM to run a SELECT statement on a workspace and receive structured results.
• Rapid Prototyping – spin up a starter workspace with a single natural‑language command, load test data, and start analysis.
• Automation – schedule recurring notebook‑based ETL jobs via chat without writing deployment scripts.
• Team Collaboration – create and share notebooks directly from an LLM conversation, then run them as jobs.
• Multi‑tenant Management – switch between organizations interactively when working across several SingleStore accounts.

FAQ

Q: Do I need to set any environment variables? A: For the standard local setup no. Authentication is handled through browser‑based OAuth. Only Docker deployments require MCP_API_KEY.

Q: Which Python version is required? A: Python 3.10 or newer (development uses 3.11).

Q: Can I use the server with other MCP‑compatible tools? A: Yes. The server follows the Model Context Protocol, so any client that supports MCP can connect.

Q: How do I see the list of available tools? A: Type /help in the chat window of your MCP client.

Q: Is the server open‑source? A: Yes, the code is available under the MIT License on GitHub.

SingleStore MCP Server

Model Context Protocol (MCP) is a standardized protocol designed to manage context between large language models (LLMs) and external systems. This repository provides an installer and an MCP Server for Singlestore, enabling seamless integration.

With MCP, you can use Claude Desktop, Claude Code, Cursor, or any compatible MCP client to interact with SingleStore using natural language, making it easier to perform complex operations effortlessly.

💡 Pro Tip: Not sure what the MCP server can do? Just call the /help prompt in your chat!

Requirements

• Python >= v3.10.0
• uvx installed on your python environment
• VS Code, Cursor, Windsurf, Claude Desktop, Claude Code, Goose or any other MCP client

Getting started

Getting started

First, install the SingleStore MCP server with your client.

Standard config works in most of the tools:

{
  "mcpServers": {
    "singlestore-mcp-server": {
      "command": "uvx",
      "args": [
        "singlestore-mcp-server",
        "start"
      ]
    }
  }
}

No API keys, tokens, or environment variables required! The server automatically handles authentication via browser OAuth when started.

Automatic setup:

uvx singlestore-mcp-server init --client=claude-desktop

Manual setup: Follow the MCP install guide, use the standard config above.

Automatic setup:

uvx singlestore-mcp-server init --client=claude-code

This will automatically run the Claude CLI command for you.

Manual setup:

claude mcp add singlestore-mcp-server uvx singlestore-mcp-server start

Automatic setup:

uvx singlestore-mcp-server init --client=cursor

Manual setup: Go to Cursor Settings -> MCP -> Add new MCP Server. Name to your liking, use command type with the command uvx singlestore-mcp-server start. You can also verify config or add command line arguments via clicking Edit.

Automatic setup:

uvx singlestore-mcp-server init --client=vscode

Manual setup: Follow the MCP install guide, use the standard config above. You can also install using the VS Code CLI:

code --add-mcp '{"name":"singlestore-mcp-server","command":"uvx","args":["singlestore-mcp-server","start"]}'

After installation, the SingleStore MCP server will be available for use with your GitHub Copilot agent in VS Code.

Automatic setup:

uvx singlestore-mcp-server init --client=windsurf

Manual setup: Follow Windsurf MCP documentation. Use the standard config above.

Automatic setup:

uvx singlestore-mcp-server init --client=gemini

Manual setup: Follow the MCP install guide, use the standard config above.

Automatic setup:

uvx singlestore-mcp-server init --client=lm-studio

Manual setup: Go to Program in the right sidebar -> Install -> Edit mcp.json. Use the standard config above.

Manual setup only: Go to Advanced settings -> Extensions -> Add custom extension. Name to your liking, use type STDIO, and set the command to uvx singlestore-mcp-server start. Click "Add Extension".

Manual setup only: Open Qodo Gen chat panel in VSCode or IntelliJ → Connect more tools → + Add new MCP → Paste the standard config above.

Click Save.

Using Docker

NOTE: An API key is required when using Docker because the OAuth flow isn't supported for servers running in Docker containers.

{
  "mcpServers": {
    "singlestore-mcp-server": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm", "--init", "--pull=always",
        "-e", "MCP_API_KEY=your_api_key_here",
        "singlestore/mcp-server-singlestore"
      ]
    }
  }
}

You can build the Docker image yourself:

docker build -t singlestore/mcp-server-singlestore .

For better security, we recommend using Docker Desktop to configure the SingleStore MCP server—see this blog post for details on Docker's new MCP Catalog.

Components

Tools

The server implements the following tools:

• get_user_info: Retrieve details about the current user

  • No arguments required
  • Returns user information and details

• organization_info: Retrieve details about the user's current organization

  • No arguments required
  • Returns details of the organization

• choose_organization: Choose from available organizations (only available when API key environment variable is not set)

  • No arguments required
  • Returns a list of available organizations to choose from

• set_organization: Set the active organization (only available when API key environment variable is not set)

  • Arguments: organization_id (string)
  • Sets the specified organization as active

• workspace_groups_info: Retrieve details about the workspace groups accessible to the user

  • No arguments required
  • Returns details of the workspace groups

• workspaces_info: Retrieve details about the workspaces in a specific workspace group

  • Arguments: workspace_group_id (string)
  • Returns details of the workspaces

• resume_workspace: Resume a suspended workspace

  • Arguments: workspace_id (string)
  • Resumes the specified workspace

• list_starter_workspaces: List all starter workspaces accessible to the user

  • No arguments required
  • Returns details of available starter workspaces

• create_starter_workspace: Create a new starter workspace

  • Arguments: workspace configuration parameters
  • Returns details of the created starter workspace

• terminate_starter_workspace: Terminate an existing starter workspace

  • Arguments: workspace_id (string)
  • Terminates the specified starter workspace

• list_regions: Retrieve a list of all regions that support workspaces

  • No arguments required
  • Returns a list of available regions

• list_sharedtier_regions: Retrieve a list of shared tier regions

  • No arguments required
  • Returns a list of shared tier regions

• run_sql: Execute SQL operations on a connected workspace

  • Arguments: workspace_id, database, sql_query, and connection parameters
  • Returns the results of the SQL query in a structured format

• create_notebook_file: Create a new notebook file in SingleStore Spaces

  • Arguments: notebook_name, content (optional)
  • Returns details of the created notebook

• upload_notebook_file: Upload a notebook file to SingleStore Spaces

  • Arguments: file_path, notebook_name
  • Returns details of the uploaded notebook

• create_job_from_notebook: Create a scheduled job from a notebook

  • Arguments: job configuration including notebook_path, schedule_mode, etc.
  • Returns details of the created job

• get_job: Retrieve details of an existing job

  • Arguments: job_id (string)
  • Returns details of the specified job

• delete_job: Delete an existing job

  • Arguments: job_id (string)
  • Deletes the specified job

Note: Organization management tools (choose_organization and set_organization) are only available when the API key environment variable is not set, allowing for interactive organization selection during OAuth authentication.

Development

Prerequisites

• Python >= 3.11
• uv for dependency management

Setup

1. Clone the repository:

git clone https://github.com/singlestore-labs/mcp-server-singlestore.git
cd mcp-server-singlestore

1. Install dependencies:

uv sync --dev

1. Set up pre-commit hooks (optional but recommended):

uv run pre-commit install

Development Workflow

# Quick quality checks (fast feedback)
./scripts/check.sh

# Run tests independently
./scripts/test.sh

# Comprehensive validation (before PRs)
./scripts/check-all.sh

# Create and publish releases
./scripts/release.sh

Running Tests

# Run test suite with coverage
./scripts/test.sh

# Or use pytest directly
uv run pytest
uv run pytest --cov=src --cov-report=html

Code Quality

We use Ruff for both linting and formatting:

# Format code
uv run ruff format src/ tests/

# Lint code
uv run ruff check src/ tests/

# Lint and fix issues automatically
uv run ruff check --fix src/ tests/

Release Process

Releases are managed through git tags and automated PyPI publication:

1. Create release: ./scripts/release.sh (interactive tool)
2. Automatic publication: Triggered by pushing version tags
3. No manual PyPI uploads - fully automated pipeline

See scripts/dev-workflow.md for detailed workflow documentation.