SECURITY BY DESIGN

Useful AI work should never require a leap of faith.

FleetRun is designed for a governed pilot: least-privilege connections, visible approvals, and a record of what happened.

Fleet isolation

Every Fleet is scoped independently. Access is enforced in the database, not by an interface convention.

Encrypted connections

OAuth credentials are encrypted before storage and are never passed to a model or browser.

Approval gates

Email drafts, calendar changes, and other consequential work wait for an authorised human decision.

Traceable work

Missions capture model activity, approved actions, tool usage, and cost so work can be reviewed.

Before you connect

The questions we expect a careful operator to ask.

What can the agent read?

Only the tools and scopes configured for the Fleet. Start with constrained, read-only access where possible.

What can it send or change?

Consequential actions are approval-gated in the pilot. The system should make the boundary visible before work begins.

What will it cost?

A mission reserves its estimated maximum before execution and records actual usage in the mission trace.

Pilot scope, stated plainly

FleetRun is not presented as a certified compliance program. Before enterprise claims, we will complete the relevant independent review, incident-response process, and compliance work.

Questions about data handling? Contact us.