What is a check before an agent acts?
A check is a decision that happens before a tool call or a write goes out. The model can propose. Nexus decides. Allow, stop, or a person. Then it writes the record.
Last updated 2026-08-17
Why does the check matter?
The risk is not a weak model. It is an action that leaves without a check.
Agents propose tool calls and writes. If those leave unchecked, you cannot take them back. The check sits on the path out. If Nexus says no, the action does not happen.
How does the check work in Nexus?
Every consequential action gets a verdict: allow, stop, or a person. The verdict is written as a record you can export.
When it matters, a person sits on the path and the action waits. The decision is not a chat message. It is a record.
How is this different from a prompt rule?
A prompt rule asks the model to behave. A check decides whether the action may leave.
Prompt rules can be ignored. A check is on the path out. If Nexus says no, the action does not happen.
Frequently asked questions
What is a check before an agent acts?
A check is a decision before a tool call or a write leaves: allow, stop, or a person. Then Nexus writes the record.
Who needs a check?
Any team that lets agents act in production: send, write, move, change. Peace of mind when AI agents act.
Does Hardalion provide this?
Yes. Hardalion Nexus is where teams build and run production agents. See hardalion.com/nexus.
Related guides
What is Hardalion Nexus?
Hardalion Nexus is where teams build and run production AI agents. Before a tool call or a write goes out, Nexus enforce…
How do you keep a record when agents act?
Keep a record of the check. Allow, stop, or a person. Write the decision. Export it when you need to prove what happened…