Sources
Which sources work with Portward
PostgreSQL is available in the beta. MySQL, SQL Server and S3-compatible storage are being tested, and more are planned.
01Connectors
What you can connect today
- Works today
PostgreSQL
Every step passes our automated tests, from connecting to enforcing an excluded column. It has not yet been run against a customer's live database.
- Being tested
MySQL, Microsoft SQL Server, S3-compatible object storage
These pass tests against local copies of each system. You can't select them until they have been tried against a live account.
- Planned
Snowflake, BigQuery, Databricks, Amazon Redshift, Google Drive, OneDrive and SharePoint, web APIs, uploaded files
Built or partly built, and not yet available to select.
02Setup
What happens when you connect a source
1. Enter the connection details
The credential is encrypted before it is stored. It is never written to logs or error messages, and it is never sent to an AI tool.
2. Portward reads the structure
Schemas, tables, columns and keys that the connection is allowed to see. This step makes no AI model call.
3. You set the rules
Who may use the source, and which rows and columns are excluded. The rules belong to the source, so they apply wherever it is used.
4. You add definitions
What each table and number means. A definition your team approved is labelled Governed, and one nobody has reviewed is labelled Provisional.
03Your data
What Portward copies from a source
What Portward copies
To answer a question, Portward copies the tables that question uses into its own storage and runs the calculation there. A copy is limited to 100,000 rows per table by default, and never more than 1,000,000.
What it leaves alone
Tables no question has used are not copied. Portward never writes to your source.
Connect a PostgreSQL database and ask Claude about it
Create an account, enter the connection details, and add Portward to your AI tool.