Ask your database anything. In plain English, no SQL.

No. Every connection Datafliq opens is read-only: your data can be read, never changed or deleted. On Postgres, MySQL and SQLite the database itself rejects any INSERT, UPDATE, DELETE or DROP. Every source also passes a second check that allows only a single, straightforward read statement through, so reads reach your data and nothing else does. MongoDB connections only ever issue a find().

No, and nothing gets installed in your environment. Datafliq is hosted: your database stays where it is, and our server connects out to it using the credentials you provide, runs the query, and passes the rows back to your assistant. You will need to allow network access from our servers to your database.

It sees the rows your query returns, which is how it answers the question. It does not see tables it did not query, sources you have not connected, or anything a teammate has access to but you do not. Your database credentials are never exposed to the model.

No. Results pass straight through to your assistant and are not written to disk or retained.

An assistant does not need to load your whole schema to answer a question. It can search your schema by keyword to find the handful of relevant tables, then summarise just those. Asking it to summarise an entire 400-table source will describe every table one by one and is slow, so narrow it with a search first.

You decide per source. Members can only query the sources you explicitly assign to them, and that scoping is applied when their assistant connects, not in the interface. Admins can query every source in the organisation. Revoking access takes effect on the next query.