AgentPass
Your agent asks. A named person approves the exact action on their phone.
AgentPass sits in front of risky tool calls. A policy decides whether a call runs, stops, or waits for one or two people. They approve with a passkey, and the tool runs only with a signed decision that matches the arguments they saw.
Request a 30-day pilotHow it works
- The agent sends AgentPass the tool name, the exact arguments, a risk level and a one-line summary.
- Your rules decide. For example, payments above a threshold need two people, email to a domain outside your list needs one, and anything named delete_* is refused.
- The approver sees the arguments as they will run and confirms with a fingerprint, face or device PIN.
- The agent receives a signed decision. Your tool executor checks the signature and the arguments before running. If the IBAN changed after approval, the check fails.
What it covers
| Different arguments | A decision issued for one IBAN is refused for another. |
|---|---|
| Different tool or agent | A decision issued for one tool is refused for another. |
| Reuse and expiry | One approval works once, and expires. If nobody answers in time, the action is denied. |
| Pressure option | An approver under pressure can press a button that looks like an ordinary refusal and silently alerts administrators. |
| Audit | Every request, decision, rule change and sign-in goes into a hash-chained log with CSV export. |
Write to us with a line about your system or your team. We reply from a named person within two working days.
Request a 30-day pilotCovered by our client terms. Client terms