> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# How do I use a feature store with Kumo?

> Include existing features as graph inputs

Export the features to a supported data source using [Linux](/kumo-engine/user-code-guide/connect-your-data-linux), [Snowflake](/kumo-engine/user-code-guide/connect-your-data-snowflake), or [Databricks](/kumo-engine/user-code-guide/connect-your-data-databricks), create a table with the entity keys and feature columns, and connect it to your graph. A feature store is optional; Kumo Engine also learns features from source tables and their relationships.

For temporal tasks, preserve the feature values available at each historical prediction time. Joining today's feature values into historical examples can leak future information.

See [Tables](/kumo-engine/user-code-guide/set-up-a-graph/tables) and [Graphs](/kumo-engine/user-code-guide/set-up-a-graph/graphs) for setup.