- The data plane accepts Postgres connections and checks every statement against the caller’s credentials, role, and limits before DuckDB runs it.
- The control plane creates databases and users, and mints short-lived credentials on demand.
- Your storage holds each database’s catalog in Postgres and its tables as Parquet files in object storage, isolated from every other database.
Airhouse checks every statement before DuckDB runs it, can hand out short-lived credentials instead of stored passwords, and keeps each database's catalog and files isolated in open formats on infrastructure you control.
Why Airhouse
- Works with every Postgres tool.
psql, JDBC,psycopg2, DBeaver, Tableau, and agents connect over the standard wire protocol. - Analytical speed. DuckDB runs joins, window functions, and aggregates over large tables quickly, reading only the files and columns a query needs.
- Open data, your infrastructure. Tables are standard Parquet files in your object storage with a DuckLake catalog in your Postgres, so the data stays readable without Airhouse.
- Isolation by default. Each database has its own catalog, storage location, users, and query limits.
- No long-lived keys. Service accounts mint short-lived credentials on demand, so a platform never stores a password per user.
Next steps
- Getting access: how to get Airhouse from Oxygen.
- Quickstart: run Airhouse locally and run your first queries.
- Architecture: key concepts, the life of a query, and deployment topologies.
- Connect: psql, drivers, and BI tools.