Security note

Last reviewed September 2026

Runbook Labs agents work inside your finance systems, so we treat access, data handling and approvals as product features rather than paperwork. This note describes how a deployment is set up and what you can expect to see.

Access

• Connections use your identity provider through single sign-on where the system supports it.
• Each runbook declares the systems it reads and writes. Read-only connections are the default.
• Permissions are role based. Approvers, reviewers and administrators are separate roles.

Data handling

• Data is encrypted in transit and at rest.
• Runs are scoped to the entities and periods you connect. Nothing else is read.
• Your ledger data is never used to train models.
• Retention windows for run logs and exported files are set per customer.

Approvals

• Steps you mark as sensitive stop and wait for a named approver before anything posts.
• Approval requests carry the agent's reasoning and the source records it used.
• Thresholds for automatic posting are set in the runbook and can be changed at any time.

Audit trail

• Every match, hold, posting and approval is written to an append-only log.
• Each journal entry links back to the source records it was built from.
• Logs can be exported in full for internal or external audit.

Questions

Security reviews, questionnaires and data processing terms are handled on the first call. Ask for the full documentation set when you book one.