Security embedded in every agent.

Every deployed agent runs on its own dedicated machine with its own filesystem and network address. Nothing is shared between one agent and another.

No. Connected-account credentials stay behind a gateway the agent cannot read. Secrets are swapped in the moment a tool runs and stripped from every result and error the model gets back.

Secrets and connected-account tokens are sealed with AES-256-GCM before they ever reach the database, and all traffic runs over HTTPS and TLS.

Yes. Mark any tool as approval-required and the agent pauses mid-run, waiting for a person to approve or deny before it acts.

Yes. Every tool call an agent makes is captured in a trace, with secret values redacted before export, so you can review exactly what happened and when.

Each agent carries a token you can rotate at any time, and workspace roles decide who is allowed to use it.

Each agent runs on its own isolated cloud machine with a dedicated storage volume, so its files persist across restarts and stay separate from every other agent.

Yes. Talona is GDPR compliant, and a Data Processing Addendum is available on request.

Yes. Talona is HIPAA compliant for handling protected health information.

Yes. Talona is PCI DSS compliant.

Talona is aligned with SOC 2 and ISO 27001. Both are designed into how the platform runs, though Talona is not yet formally certified against either.

Talona is aligned with the Data Privacy Framework. A formal certification is not yet in place.