Trust & security
Built to be read by your security team.
DataConverse runs on our managed cloud or inside your own infrastructure. Either way it connects read-only, indexes your schema rather than your data, and shows every query it writes. Here is exactly what that means, in writing.
Two ways to run it.
Managed cloud
We run it for you. Nothing to install on your side.
- We connect from a fixed set of service IP addresses that you whitelist. Anything else is refused at your firewall, by you.
- Every database is registered with a read-only credential that you create.
- Your schema is indexed in an isolated index per database. Your rows stay in your database.
On premise
It runs inside your own infrastructure, with our team supporting it.
- Installed inside your network. Data and models stay within your perimeter.
- Bring your own model, hosted inside your perimeter.
- The same read-only access, schema indexing and controls as the managed cloud.
Commitments your reviewers can check.
Read-only, always
Every database is connected with a read-only credential. The product does not issue writes, and a read-only role means it cannot, whatever happens.
We index your schema, not your data
Extraction reads metadata only: tables, columns, types, keys and relationships. The AI writes each query from that schema, and your rows are never copied into our index.
Isolated per database
Each database gets its own vector index. Nothing is pooled across your databases, and nothing is shared between customers.
Role-based access
Admins add members and grant access database by database. A person can only query the databases they have been granted.
Every query is visible and checked
Generated queries are sanitised and validated for execution safety before they run, and shown to the person who asked, every time.
Supported to an SLA
After go-live the deployment moves to active production support, with a service level rather than best effort.