How do I connect my warehouse to Claude?

Add Rig as an MCP server in Claude. Rig connects to your warehouse, builds the context layer, and Claude can then answer data questions with governed SQL.

Does Claude see my raw data?

Claude works through Rig's governed context layer and runs validated queries against your warehouse. It sees your schema and metric definitions, and returns query results, not bulk exports.

Will Claude give consistent numbers?

Yes. Rig answers from agreed metric definitions and validates each query in a sandbox, so the same question returns the same number across the team.

Which warehouses work with Claude through Rig?

Snowflake, BigQuery, Redshift, Postgres and DuckDB/MotherDuck are all supported, along with SaaS sources synced by Rig Ingest.

AI tools Supported

Claude + Rig

The easiest way to query your warehouse data in Claude

Claude is Anthropic's AI assistant, available in the Claude apps and on the web, that can connect to external tools and data over MCP.

Try Rig Book a demo

What Rig gives Claude

Once connected, Claude can reason over your governed data through Rig:

Warehouse tables and views
Metric definitions
Join paths
dbt lineage
Role-based access
Sandbox-validated SQL

How Claude connects to Rig

Rig runs as an MCP server that Claude connects to. Once added, Claude can see your governed schema and metric catalog and answer data questions by generating validated SQL against your warehouse, never raw dumps of your tables.

Answer data questions in Claude without writing SQL

With Rig connected, Claude turns a plain-English question into validated warehouse SQL using your real schema and metric definitions. A non-technical operator gets the same number an analyst would, because Claude reasons over the modelled context rather than guessing.

Ask Rig

What was net revenue retention last quarter?
Show monthly active accounts by plan
Which segments are churning fastest and why?

Keep Claude's answers consistent and trustworthy

Rig validates every query Claude generates in a sandbox before it runs, and answers come from agreed metric definitions. That ends the ten-people-ten-numbers problem where the same question returns different figures depending on who asks.

Ask Rig

How is active customer defined here?
Recompute that with the official revenue metric
Show the SQL behind this number

Go from a number in Claude to an action

Rig can carry an answer into a workflow, so Claude does not stop at the insight. From a single thread you can find the at-risk accounts, then push an update to your CRM or a Slack alert, governed the whole way.

Ask Rig

List accounts with declining usage and flag them in the CRM
Draft a summary of this churn analysis for the team
Alert the owner of each at-risk account in Slack

Let Claude trace a metric to its source

Because Rig carries dbt lineage and join paths, Claude can explain how a number is built, from the metric down to the warehouse table. That makes answers auditable rather than a black box.

Ask Rig

What feeds the revenue figure you just gave me?
Walk through the lineage for monthly recurring revenue
Which tables and joins did that query use?

Frequently asked questions

How do I connect my warehouse to Claude?: Add Rig as an MCP server in Claude. Rig connects to your warehouse, builds the context layer, and Claude can then answer data questions with governed SQL.
Does Claude see my raw data?: Claude works through Rig's governed context layer and runs validated queries against your warehouse. It sees your schema and metric definitions, and returns query results, not bulk exports.
Will Claude give consistent numbers?: Yes. Rig answers from agreed metric definitions and validates each query in a sandbox, so the same question returns the same number across the team.
Which warehouses work with Claude through Rig?: Snowflake, BigQuery, Redshift, Postgres and DuckDB/MotherDuck are all supported, along with SaaS sources synced by Rig Ingest.