What your agents can do — and what needs approval
See what the system may read, write, and do, and what pauses for your sign-off.
Security in Niyatna is not a disclaimer you scroll past. It is a visible set of choices about what your agents may read, write, schedule, send, and report.
You decide those limits up front, and Niyatna makes them clear before work starts and easy to check before anything risky goes out.
The limits you set
You can set limits around:
- People — who may ask for or approve this work.
- Information — which documents, notes, and records may be used.
- Tools — which outside systems may be called.
- Actions — what may be done, held for review, or refused.
- Proof — what evidence is needed before the result counts.
The three kinds of action
| Action | What it means | Examples |
|---|---|---|
| Allowed | Low-risk work can go ahead inside the limits you set. | Looking things up, drafting, formatting, preparing a report. |
| Needs review | The work can be prepared, but it waits for you before it is finished. | Sending an outside message, publishing, changing data, paying. |
| Refused | The request is outside what your company allows. | Using credentials you did not approve, touching out-of-scope data, sharing something you did not authorise. |
The rule
A good setup makes the next risky step obvious. If it is unclear who asked, what it touches, which tool it uses, or what proof it needs, Niyatna pauses and asks instead of guessing.