Introduction to Atlas

Ask questions of your data in plain English. Atlas is the trust layer between your business questions and your data — SQL warehouses or REST/OpenAPI services — agent-native, YAML-defined, deploy-anywhere.

Atlas is the AI data analyst you can run anywhere. It answers plain-English questions across your SQL warehouses and REST APIs, grounded in a YAML semantic layer that humans author and AI agents consume.

Pick your path

Atlas works across four onboarding shapes. Pick the one that matches what you want to do — each card is a single CTA, and you can ignore the rest of the page until you've completed it.

The semantic layer is the product

Everything else — the agent loop, the MCP server, the chat widget, the validation pipelines for SQL and REST/OpenAPI — exists to make the YAML you write the source of truth for how questions become answers. Every field in the format exists because an LLM needs it: sample_values to ground the agent, glossary.status: ambiguous to force clarifying questions, metrics.objective to pick MAX vs MIN, query_patterns to teach canonical join shapes.

The fastest way to feel why this works: try the hosted demo and ask one of these canonical NovaMart questions, then read the SQL the agent wrote:

• "What's our GMV this quarter?"
• "What's our top-performing category by GMV this month?"
• "Show me revenue last quarter." — the agent asks which definition (GMV vs net revenue vs seller revenue) because revenue is status: ambiguous in the glossary
• "What are our most common return reasons?"

---

Go deeper

What's new

• Changelog — Atlas release notes