Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.summand.com/llms.txt

Use this file to discover all available pages before exploring further.

Summand is a hosted analysis platform for structured data. Point it at a CSV or a database table and it builds a rich, interpretable model of the data — feature importances, statistics, embeddings, and human-readable descriptions — then layers a set of product features on top so you can actually use them.

​
What Summand does

Every dataset gets the same treatment automatically:
  • Ingestion. Upload a CSV or connect Snowflake or Delta Sharing.
  • Semantic layer. Summand trains an Explainable Boosting Machine, computes column statistics, generates feature descriptions with Claude, and produces a 2D embedding. See Semantic layer overview.
  • Surprise Finder. Once the model is trained, Summand looks for the patterns it didn’t expect and surfaces them as ranked findings.
  • Ask Summand. A chat panel lets you query the dataset in natural language; SQL, charts, and explanations stream back in-line.
  • Reports. Compose a branded PDF from blocks — charts, surprises, feature importances, narrative text — and export or share it.

​
The features

Datasets

The unit of analysis. Visibility, sharing, and where every other feature lives.

Connectors

Snowflake, Delta Sharing, and CSV uploads — with scheduled refresh.

Ask Summand

Natural-language chat grounded in your dataset’s semantic layer.

Surprise Finder

ML-generated findings: where your data deviates from the model.

Reports

Block-based editor with one-click PDF export.

Semantic layer

The artifacts that power every other feature.

​
Build on top of Summand

Everything you can do in the dashboard is also available programmatically. Generate an API key, or wire your AI assistant directly into your datasets over MCP.

Development tools

API keys, REST API, and the Summand MCP server.