Skip to main content

DuckDB as a service: what Quack on Demand adds

DuckDB is an embedded analytical engine: it runs inside your process, trusts whoever opened it, and has no network listener, no users and no grants. Quack on Demand keeps DuckDB as the engine and adds the service layer around it, so many people and tools can query shared DuckDB and DuckLake data with identity, authorization and scale.

What a single DuckDB does not give you​

You needA single DuckDBBehind Quack on Demand
A server other machines can reachNo listenerArrow Flight SQL on :31338 and native Quack on :9494
Users and passwordsNoneBuilt-in accounts, JWT, OAuth and OIDC providers
Single sign-onNoneKeycloak, Google, Azure AD and Cognito for SQL clients; any OIDC IdP including Okta for the admin UI; SCIM
Grants, row filters, column masksNoneRoles, table verbs, row-level policies, column masking
Several teams on one deploymentOne file, one processTenants, databases and pools, isolated at the edge
JDBC, ODBC, ADBC driversLocal bindings onlyStandard Flight SQL drivers for BI tools and notebooks
More than one machineOne processPools of DuckDB nodes that scale out and suspend to zero

How to read this section​

Each page below answers one question in the words people search for, shows the mechanism with one example, and links into the operator and concept pages for the full detail.

Quack on Demand runs as a single Docker container on one node, or on Kubernetes for multi-host fleets. Start with the Quickstart, or read the Architecture for how the pieces fit.