Bring your own warehouse

Kubit queries traces and events where they already live, in your Snowflake, BigQuery, Databricks, ClickHouse or Redshift. Nothing is copied out of your account. No data silo to secure. Traces retained in your control.

How Kubit connects to your warehouse

Kubit connects to your warehouse with a dedicated, read-only identity that you create for it. Every setup page follows the same pattern:

Part	What you create	Why
Identity	A Kubit user, role, service account or service principal	Kubit signs in as this identity and nothing else
Read access	`SELECT` on the schemas, tables or views Kubit should query	Kubit only reads the data you grant
Compute	A warehouse sized for Kubit, on Snowflake and Databricks	Kubit’s queries run on compute you control and can monitor
Auxiliary database	A small database Kubit owns, on Snowflake (`KUBIT`) and ClickHouse (`KUBIT_DB`)	Kubit keeps auxiliary tables and views there to support some features
Network access	Kubit’s IP ranges in your firewall, or VPC access on Redshift	Only Kubit can reach the warehouse from outside your network

Supported warehouses

Warehouse	Kubit signs in with	Setup page
Snowflake	A `KUBIT` user with key pair auth	Connect Snowflake
BigQuery	A `Kubit` service account with a JSON key	Connect BigQuery
Databricks	A service principal on a Serverless SQL warehouse	Connect Databricks
ClickHouse	A `KUBIT` user with `KUBIT_R` and `KUBIT_W` roles	Connect ClickHouse
Redshift	A `kubit` user, reached through VPC access	Connect Amazon Redshift

For agent traces, the Warehouse origin in Create Workspace lists Databricks, Snowflake, BigQuery and ClickHouse. Set up your organization walks through creating that workspace.

Your warehouse or Kubit compute

A workspace reads data from your own warehouse or from a Kubit-hosted one.

Keep Kubit’s queries fast and cheap

Kubit’s queries filter on time and event name, so the same two choices help on every warehouse.

Practice	Where it’s covered
Cluster, partition or sort fact tables by event date, then by event name	Snowflake, Databricks, Redshift
Start with the smallest compute and size up only when query history shows spilling or queuing	Snowflake, Databricks
Use a separate identity and compute for development	The optional last step of each setup page

Allow Kubit’s IP ranges: the current ranges and how to keep firewall rules up to date.
Set up your organization: create a workspace on your warehouse.

Start here

Agent Observability

Product Analytics