SingleStore MCP server
SingleStore's official MCP server lets agents run SQL on workspaces, manage starter workspaces and Stage files, and orchestrate notebooks and jobs.
The SingleStore MCP server is the official integration from SingleStore Labs that connects an AI agent to SingleStore Helios, the distributed SQL database built for fast analytics and real-time applications. Its central tool, run_sql, executes SQL against a connected workspace so an agent can query and analyze data directly, while a set of management tools let it inspect the account: get the current user and organization, choose and set the active organization, and list workspace groups, workspaces, and regions. The agent can also provision and operate compute — create, resume, and terminate starter (virtual) workspaces — making it practical to spin up an environment, run analysis, and tear it down without leaving the conversation.
Beyond querying, the server exposes SingleStore's data-application surface: it can create and upload Jupyter notebooks into SingleStore Spaces, schedule jobs from a notebook, and inspect, fetch, and delete those jobs, plus manage files in Stage (list, get, upload, create folders, move, and delete), the platform's object store. It is published on PyPI as singlestore-mcp-server and is typically launched with uvx (uvx singlestore-mcp-server start); when no API key is set it runs an automatic browser-based OAuth flow, and for headless or Docker deployments it authenticates with an MCP_API_KEY. That makes it a strong backend for analytics copilots and data-app automation on SingleStore.
Quick install
Copy-paste configs are provided for all 8 supported clients. Pick your client below.
Available tools
| Tool | Description |
|---|---|
| get_user_info | Retrieves details about the current user. |
| organization_info | Retrieves details about the user's current organization. |
| choose_organization | Lists the organizations available to the user (OAuth sessions only). |
| set_organization | Sets the active organization (OAuth sessions only). |
| workspace_groups_info | Returns details about the workspace groups the user can access. |
| workspaces_info | Returns details about the workspaces in a workspace group. |
| resume_workspace | Resumes a suspended workspace. |
| list_starter_workspaces | Lists the starter (virtual) workspaces available to the user. |
| create_starter_workspace | Creates a new starter workspace using the SingleStore SDK. |
| terminate_starter_workspace | Terminates a starter workspace. |
| list_regions | Lists the regions available for creating workspaces. |
| list_sharedtier_regions | Lists the regions available for shared-tier (starter) workspaces. |
| run_sql | Executes a SQL query against the connected SingleStore workspace. |
| create_notebook_file | Creates a Jupyter notebook file in SingleStore format. |
| upload_notebook_file | Uploads a notebook file into SingleStore Spaces. |
| create_job_from_notebook | Creates a scheduled job that runs a notebook. |
| get_job | Retrieves the details of a scheduled job by ID. |
| delete_job | Deletes a scheduled job by ID. |
| stage_list_files | Lists files and folders in a Stage. |
| stage_get_file | Retrieves a file from a Stage. |
| stage_create_folder | Creates a folder in a Stage. |
| stage_upload_file | Uploads a file to a Stage. |
| stage_move | Moves or renames a file or folder in a Stage. |
| stage_delete | Deletes a file or folder from a Stage. |
Required configuration
- MCP_API_KEYOptional
SingleStore API key. Optional for interactive OAuth use; required for Docker/headless deployments.
What you can do with it
Analytics copilot over SingleStore
Let an agent list workspaces, connect to one, and run run_sql to answer analytics questions over real-time data, surfacing results without anyone opening the SingleStore portal.
Automate notebooks and jobs
Have the agent create and upload a notebook into SingleStore Spaces, schedule it as a job with create_job_from_notebook, and check on it later with get_job — turning a one-off analysis into a repeatable pipeline.
FAQ
- Is it free?
- The MCP server is open source under the MIT license and free to run. You pay only for the SingleStore Helios workspaces and compute it connects to or provisions.
- Does it support remote/OAuth?
- Yes. Run locally with uvx, and when no API key is set the server performs an automatic browser-based OAuth flow to authenticate. For Docker or headless deployments, set MCP_API_KEY with a SingleStore API key instead.
- Can it create and delete infrastructure?
- Yes. Alongside running SQL it can create and terminate starter workspaces and create or delete scheduled jobs and Stage files, so scope the authenticated account's permissions to what the agent should be allowed to do.