Your own database
Each customer receives a separate PostgreSQL database. Customer business data is not stored in a shared table with a tenant identifier.
Security and data residency
Fratera separates customer data at database level and enforces access inside the database. EU and US deployment regions are available, with the same core security architecture in every edition.
Security by design
Contract repositories contain commercial terms, personal data and documents that should never cross a customer boundary. Fratera therefore gives every customer a separate PostgreSQL database rather than separating tenants with an application filter inside one shared database.
Inside that database, row-level security applies the signed-in user’s permissions to direct data access as well as the application. Physical separation and database authorization solve different problems; Fratera uses both.
Each customer receives a separate PostgreSQL database. Customer business data is not stored in a shared table with a tenant identifier.
Your region is selected when the workspace is created. The customer database, document storage and regional application services are deployed for that region.
Data is encrypted at rest using AES-256 and protected in transit with TLS 1.2 or later. Database connections and document storage are encrypted.
Compose access from reader, contributor, approver, procurement and administrator capabilities. Restricted visibility groups protect sensitive records, while delegation-of-authority rules control approval thresholds.
Use email sign-in or OpenID Connect SSO. Microsoft Entra integration and SSO-only enforcement are being completed for launch. Roles remain controlled inside Fratera; directory sync through SCIM is planned, not presented as a current capability.
Approval decisions, access changes and key contract and vendor record changes are written to the audit trail. Administrators can export customer data, configure retention and manage deletion at the end of the relationship.
Row-level security is default-deny and evaluates requests as the signed-in user. The same authorization boundary applies when data is accessed through the interface or through direct API requests.
AI with boundaries
Fratera keeps model output outside authorization, workflow, money and record-changing decisions. Read-only questions run through controlled product tools under the signed-in user’s existing permissions.
Contract answers are bounded to documents the user can already access and include source-page evidence so the answer can be checked.
Fratera does not use submitted customer content to train models.