Verify who an agent is, who it acts for and what it’s allowed to do, before it does anything.
68%
of organisations can’t clearly tell AI agents apart from humans
74%
say AI agents often get more access than they need
Agent registry
Procurement Assistant
did:key:z6Mkq4…7Jx2
Support Triage
did:key:z6MkfT…2Qa9
Travel Booker
did:key:z6MkpL…9cW1
Report Exporter
did:key:z6Mkv8…4nR6
Code Reviewer
did:key:z6MkhS…1eZ3
Trusted by teams at
The problem
Illustration: hundreds of agents shown as small dot grids. A scan passes across them and only a handful light up as identified.
65%
of organisations had an AI agent security incident in the past year.
195M
AI agent requests a day, on just one bot-protection network.
4 in 5
organisations found AI agents running that they didn’t know about.
The framework
Who the agent is. A signed identity that can’t be borrowed or faked.
Who it acts for. Every agent tied to a named person or company.
What it’s allowed to do. Exact scopes, limits and an expiry date.
How it works
01
Give it a name, an owner and a purpose.
02
Confirm what it is and who it acts for.
03
Scoped permissions, an expiry date and a signature.
04
See every action. Pull access in real time.
The agent passport
Acts for: [Company name]
Permissions
Recent activity
Use cases
Shopping agents that buy within limits a customer sets.
Know which agent moved money, and for whom.
Agents that act for customers, with the scope to prove it.
Every internal agent registered, scoped and revocable.