Try Opteryx

Integrations

These pages show how to use Opteryx from tools you may already run. Each one covers installing the connector, the exact connection settings, a working example, and the limitations we found when testing it.

Opteryx comes in two forms, and a tool can use one or both:

  • Opteryx Core is the embedded engine. It runs inside a Python process and queries files where they live, with no server or account. Python notebooks and apps can use it directly.
  • opteryx.app is the hosted service. Tools reach it through the opteryx-sqlalchemy dialect, the OData service or Arrow Flight SQL.

Supported Tools

Tool What it's for Opteryx Core opteryx.app connection
Apache Superset Shared BI dashboards and SQL exploration — opteryx-sqlalchemy
Grafana Time-series dashboards — OData, through the Infinity plugin
Jupyter Interactive analysis in notebooks Yes opteryx-sqlalchemy, with pandas or %sql
marimo Reactive notebooks and small apps Yes opteryx-sqlalchemy, with mo.sql
Streamlit Data apps written in Python Yes opteryx-sqlalchemy, with st.connection

Power BI and Excel connect through OData: see Iceberg in Power BI and Excel. Claude and other AI assistants connect through MCP: see Using Opteryx from your assistant.

Not Supported Yet

  • dbt needs a dbt adapter, and there isn't one for Opteryx. The SQLAlchemy dialect and the Flight SQL endpoint can't stand in for one: dbt doesn't use SQLAlchemy, and both are read-only.
  • Metabase has no driver that reaches Opteryx. opteryx.app has no PostgreSQL-compatible endpoint, and Metabase doesn't read OData.

Choosing a Connection

  • SQL tools (Superset, notebooks, Streamlit) use opteryx-sqlalchemy. It needs a client ID and client secret from the Authentication API.
  • Tools that read JSON over HTTP (Grafana, Power BI, Excel) use the OData service. The demo workspace is readable without credentials. Other workspaces accept HTTP Basic authentication with the client ID and secret.
  • Arrow-native tools can use Flight SQL, which streams results as Arrow record batches.

All of these are read-only, and all of them only see the datasets your credential's policies grant. See Access and Permissions.