mcp -singlestore

This MCP server lets you interact with SingleStore using natural language via MCP clients like Claude Desktop, Claude Code, Cursor, VS Code Copilot, Windsurf, Gemini CLI, LM Studio, and more. The server exposes a set of tools for managing user info, organizations, workspaces, and SingleStore resources, enabling you to perform tasks such as retrieving user and organization details, listing workspaces and regions, running SQL queries, and handling notebooks within SingleStore Spaces. The standard config uses uvx to launch the server, and the server will guide you through authentication via browser OAuth when started. Use the /help prompt in your chat to discover available capabilities and how to invoke each tool.

Prerequisites:

• Python >= 3.10
• uvx installed in your Python environment (for launching the MCP server)
• Access to the internet to install the package from PyPI

Installation steps:

1. Create and activate a Python virtual environment (optional but recommended): python -m venv venv source venv/bin/activate # Linux/macOS venv\Scripts\activate # Windows

2. Install the Singlestore MCP Server package from PyPI: pip install singlestore-mcp-server

3. Start the MCP server using the standard uvx command configuration (as shown in the README): uvx singlestore-mcp-server start

4. Follow the OAuth flow in your browser when prompted to authenticate. The server will be ready to accept MCP client connections once started.

Notes and tips:

• Docker usage requires an API key (MCP_API_KEY) due to OAuth flow restrictions in containers. If you run the server via Docker, provide MCP_API_KEY in the environment.
• The default configuration uses the standard config with command 'uvx' and arguments 'singlestore-mcp-server' and 'start'. This works with the recommended MCP clients and tools.
• Tools exposed by the server include get_user_info, organization_info, choose_organization, set_organization, workspace_groups_info, workspaces_info, resume_workspace, list_starter_workspaces, create_starter_workspace, terminate_starter_workspace, list_regions, list_sharedtier_regions, run_sql, create_notebook_file, upload_notebook_file, and create_job_from_notebook, among others. Use /help in your chat to see current capabilities and usage examples.
• If you plan to run the server in Docker, you’ll typically build or pull a container image like singlestore/mcp-server-singlestore and pass environment variables as needed (e.g., MCP_API_KEY). See the Docker example in the README for the exact invocation.
• For automatic client integrations (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Gemini, LM Studio, Goose, Qodo Gen), follow the automatic setup instructions in each section; manual setup generally mirrors the standard config and adjusts the client as needed.