Text-to-SQL agent

Ask your data
anything_

Connect your database, auto-generate a semantic layer, and query your data in plain English. Atlas validates every query, enforces read-only access, and deploys anywhere.

Try the demoView on GitHub

Text-to-SQL Agent

AI agent that explores your schema, writes validated SQL, and interprets results. Powered by the Vercel AI SDK with multi-step reasoning.

Semantic Layer

YAML-based entity definitions with auto-profiling, glossary terms, metric definitions, and LLM enrichment. The agent reads before it writes.

Deploy Anywhere

Docker, Railway, Fly.io, Render, or Vercel. Ship a full-stack app or a headless API. One command to scaffold: bun create atlas-agent.

How it works

01

Connect your database

Point Atlas at any PostgreSQL, MySQL, or SQLite database. Read-only access enforced at every layer.

ATLAS_DATASOURCE_URL=postgresql://...
02

Generate semantic layer

Auto-profile tables, detect joins, enumerate values, generate glossary and metrics. Optionally enrich with LLM.

atlas init --enrich
03

Ask questions

Natural language in, validated SQL out. The agent reads your semantic layer before writing any query.

"What were our top 10 accounts last quarter?"

Security

Every query runs through a 4-layer validation pipeline. No data ever leaves your infrastructure.

Read-only queries
AST validation
Table whitelisting
Sandboxed execution
Statement timeouts