Cassis gives any agent current, tested, approved context on your data stack.
Every answer, SQL query, and dashboard link it returns draws on that context.
From bottom to top: a warehouse contains bronze, silver, and gold data. dbt transforms data within it. Cassis usually queries gold models and can query silver or bronze when needed. An optional semantic layer provides validated metric definitions. Cassis provides context above that layer. The context is versioned in your Git repository, and your team approves every change. Agents and visualizations consume the context. Any agent that needs data connects to Cassis through MCP, including coding agents, general-purpose assistants, and custom agents. Cassis returns any combination of answers computed on the data, SQL queries, and links to existing dashboards. Agents link users to existing dashboards, or create and update dashboards, custom apps, and artifacts from that SQL. Cassis keeps an inventory of these dashboards in its context, shown as a dashed reference line. Dashboards run their SQL through the semantic layer when present, or directly on the warehouse otherwise. Agents send feedback to Cassis. Its continuous maintenance loop detects drift, drafts fixes, runs evals, and publishes reviewed changes to the context. At the top, a user asks the agent questions and opens dashboards directly.
Any agent that needs data
Agents respond to questions, link to existing dashboards or build new ones with SQL from Cassis.
Reusable, shareable artifacts.
Governed source of truth
Continuous maintenance
Versioned in your Git repository. Your team approves every change.
Returns one or more of:
Queries through the semantic layer
Validated metric definitions
Direct warehouse connectionMetadata only. No data access.
Transformation
Model your data
Cassis runs SQL on your warehouse through the semantic layer. Without a semantic layer, Cassis provides equivalent features.
dbt transforms data in your warehouse. Cassis usually queries gold models and can access silver or bronze when needed. Cassis also works without a warehouse connection. It syncs your schema on a schedule, and your agents run the SQL it returns.
Definitions, relationships, business rules, and an inventory of your dashboards live in one context. It lives in your Git repository, and your team approves every change. Every agent and dashboard that consumes your data draws on it. Cassis queries through your semantic layer or provides equivalent semantic features itself.
Any agent can simply ask a question and get an answer. Cassis can also return the SQL behind it and links to existing dashboards. Agents use that SQL to create and update dashboards, custom apps, and artifacts. Dashboards run their SQL through the semantic layer, or directly on the warehouse.
Bring one domain and we'll go through it with you: what your agent would get wrong today, which definitions your team has never settled, and what keeping them current takes.
