Built for the day an agent misbehaves. Not the day it does not.
Agents are eager and prompts can be hostile. TapPass holds every agent to least privilege, redacts data before it moves and stops the risky calls for a person, so a bad instruction cannot become a bad outcome.
✓Security documentation on request
Six controls. On every call.
A hostile instruction can change what an agent wants. It cannot change what the agent is allowed to do.
What we can show you today. And what we will not claim.
No feature guarantees a compliance outcome. What TapPass gives your security and compliance teams is the controls and the evidence that make your own attestations defensible. The documentation behind every control is in the Trust Center.
On-premises releases are signature-checked before they run, and roll back on a failed health check.
Hosted in the EU, with a DPA on request and a clear processor boundary.
People sign in through Okta, Microsoft Entra ID or Google, under your own policies.
Is TapPass ISO 27001 or SOC 2 certified?
Not yet. ISO 27001 certification is in progress, and we do not hold SOC 2. The documentation behind every control is in the Trust Center.
Can a prompt injection make an agent break the rules?
A hostile instruction can change what an agent tries to do, not what it is allowed to do. The rules sit outside the model and hold either way.
Where are our model keys kept?
In an encrypted vault, never on people’s devices. You bring your own keys and stay in control of them.
How do people sign in?
Through single sign-on with Okta, Microsoft Entra ID or Google. Approvals are signed with a passkey on the approver’s own device.
Assume agents will misbehave. Make sure it does not matter.
Connect one agent and see every call TapPass would have held, stopped or routed elsewhere. Nothing is blocked while you learn.