Connecting DuckDB to Cloud Storage, dbt & BI Tools

DuckDB’s extension system lets it reach far beyond the local filesystem — object storage, external databases, and downstream BI tools all plug in through a small, focused set of extensions.

Querying S3 / GCS / Azure Blob with httpfs

INSTALL httpfs;
LOAD httpfs;

SET s3_region = 'us-east-1';
SET s3_access_key_id = '...';
SET s3_secret_access_key = '...';

SELECT * FROM read_parquet('s3://my-bucket/sales/2026/*.parquet');

Object storage credentials can also come from the standard AWS credential chain (env vars, ~/.aws/credentials, IAM role) via CREATE SECRET, which is the recommended approach over hardcoding keys in SQL:

CREATE SECRET (
    TYPE s3,
    PROVIDER credential_chain
);

The same httpfs extension handles plain HTTPS URLs, Google Cloud Storage (gs://), and Azure Blob Storage (azure://) with the equivalent SET variables for each provider.

dbt via dbt-duckdb

DuckDB is a first-class dbt adapter through the community-maintained dbt-duckdb package, which makes it a popular choice for local dbt development and small-scale production dbt projects without standing up a warehouse:

pip install dbt-duckdb

my_project:
  target: dev
  outputs:
    dev:
      type: duckdb
      path: 'dev.duckdb'
      extensions:
        - httpfs

From there, dbt run behaves exactly as it does against Snowflake or BigQuery — models, tests, and docs all work unchanged, just against a local (or S3-backed) DuckDB file instead.

BI Tools over JDBC / ODBC

DuckDB ships official JDBC and ODBC drivers, so tools like DBeaver, Tableau, and Power BI can connect to a .duckdb file the same way they’d connect to any other SQL database — point the driver at the file path, no server address needed.

MotherDuck: Hosted DuckDB

For teams that want DuckDB’s engine with a shared, always-on, multi-user service (collaborative access, cloud storage of the database file, hybrid local/cloud query execution), MotherDuck — founded by DuckDB’s creators — layers that on top of the open-source engine. Queries still speak the same DuckDB SQL dialect; MotherDuck just adds the “always running, shared with my team” pieces that a purely embedded database doesn’t provide on its own.