August 2026

August 16–31

MySQL catalogs — query and write MySQL databases the same way as PostgreSQL. Setup
BigQuery catalogs — register a Google BigQuery project, query it as catalog.dataset.table and join it with your other catalogs; its tables and views appear on the Tables page. Setup
Iceberg on AWS Glue — register a Glue database as an Iceberg catalog, read its tables, and create and write tables in it. Setup
Invite users by email — company admins can invite teammates from Settings → Users. Users & invitations
Python and SQLAlchemy — the thunderduck-sqlalchemy package on PyPI, for pandas, Polars and any SQLAlchemy tool, plus Apache Superset support with query cancel. Python & SQLAlchemy · Superset
ChatGPT — connect Thunderduck to ChatGPT as well as Claude. Connecting Claude & ChatGPT

Catalog checks — a new catalog's credentials are tested before it is saved, its tables are loaded right after it is created, and deleting one asks for confirmation. Catalogs
Queries only connect to the catalogs they use — a query no longer fails because an unrelated catalog is unreachable.
Clearer plan errors — when a query is refused before it runs, the error says what went wrong.
Connection examples — the API token dialog shows a SQLAlchemy URL next to the JDBC one.
AI assistant runs in Query History — queries from Claude or ChatGPT show the source mcp.
Studio on small screens — the side panes collapse, and start collapsed on a phone.

Studio sometimes showed no rows for a successful query until the page was refreshed.
A failed connection to a customer database no longer writes its password to the logs.

AI assistants over MCP — connect Claude to Thunderduck to browse your catalogs and run SQL, with a cost limit on every query and exports for large results. Connecting Claude & ChatGPT · Tools
S3, Google Cloud Storage and Azure Blob Storage catalogs — register a bucket or container and read its Parquet, CSV and JSON files by path, without writing a bucket URL. Setup
Environment variables — company-wide secret values in Settings → Environment, available to dbt runs. dbt
Cancel a running query — from Studio.
Light mode — a theme toggle in the sidebar.

Queries are planned and priced before they run — each query is checked against your catalogs first, then run in a container sized from its plan. The plan is kept with the run. Limits & cost
Results of any size — results are stored and paged in Studio, the JDBC driver and MCP instead of streamed; saving a result no longer re-runs the query.
Progress while a query runs — Studio shows progress, and a running query shows as running.
Tables without a catalog name — schema.table resolves to the catalog that holds it, so tools like DataGrip work without fully qualified names. Catalogs
dbt runs in the console — pick the dbt command for each run, read the run log and results in the console, and see each model's queries in Query History under the project's name.
Catalog credentials are kept in a secrets manager — passwords and keys are never stored with the catalog or returned by the API.

A query that ran out of memory now says so instead of failing with a generic error.
CPU and memory figures were missing for some runs.
iceberg_scan, iceberg_metadata and iceberg_snapshots did not accept a catalog path.