Solutions · Security
Let agents act, and still know why every action was allowed.
When an agent touches production, you need two things: a control that stops the dangerous call, and evidence you can put in front of an incident review. Decision Horizon gives you both from the same place.
The questions you'll be asked
And where the record answers them
| Question | Where the record answers it |
|---|---|
| Which agent did this, on whose behalf? | principal_chain: the human sponsor, the agent and any sub-agents. |
| Was it allowed at that moment? | authority_t0: grants and explicit non-grants as they were then. |
| Had it read a secret first? | horizon: what it observed, and whether that content reached the call. |
| Which rule let it through, or stopped it? | policy: rule id, verdict, and who decided. |
| Has anyone edited this since? | sig and the hash chain, checked offline with dh verify. |
How it helps
What you get
Prompt injection, contained
Rules can depend on what the agent read, so a secret in view blocks a public write, in the same session or a later one.
Blind spots in writing
Every record lists what its insertion point can't see, so you know where coverage ends.
Into your SIEM
Export OCSF, CEF or OpenTelemetry spans and alert on blocked and held actions.
In a scripted prompt-injection replay, 6 of 6 exfiltration shapes were blocked and the legitimate fix still landed.
See it on one of your own workflows.
We'll put Decision Horizon in front of a real agent workflow of yours and replay the record with you.