PRANUS

Security: AI knows exactly what the person knows, nothing more

Once AI agents work inside your business system, the core security question is what they can see and what they can do on their own. PRANUS's answer: they see only what the person is allowed to see, and they change nothing without that person's approval.

The risk isn’t AI. It’s access without limits

An owner who lets AI into customer conversations, contracts, and finances wants to know two things: what the AI will see, and what it can do without asking. That’s a fair concern, and for many buyers security matters more than any feature.

PRANUS is built so both questions have short answers. AI sees only what the person is allowed to see. And it changes nothing without a person saying yes.

Permissions come from your org chart

In PRANUS, permissions aren’t assigned by hand in ten different apps. They follow from a person’s place in the TEAMUS org chart.

  • Permission packages by department. Each service has its own levels, for example viewer, operator, manager, or admin in the CRM. You can add a personal level on top.
  • Permissions propagate on their own. Move someone to another department and their access in CRMUS, CONNECTUS, MEMUS, accounting, and the messenger updates automatically. If anything drifts out of sync, reconciliation puts it back.
  • Separation of duties checks. The system warns you when one person is about to get permissions that your rules say belong to different people.
  • Offboard once, access is gone everywhere. One HR event removes permissions across all services, returns tasks to the department inbox, and revokes agent keys.

AI can’t see what the user can’t

An AI agent in PRANUS isn’t a superuser. It works with the permissions of the person it’s doing the task for.

  • Permissions before search, not after. When an agent looks for an answer in the knowledge base, documents the person can’t access are filtered out before they ever reach the model. The model simply never sees them.
  • Narrow service keys. Each agent has its own key, scoped to exactly the actions it needs. Keys can be revoked instantly.
  • Keys stay hidden from the model. Credentials for outside systems are stored separately. The model can use them but can’t read or repeat them.
  • Protection against malicious instructions in documents. If a vendor’s contract hides “ignore previous instructions,” it won’t work.

The agent suggests, a person decides

Every change an agent makes follows the same path: plan, dry run, a person’s approval, apply, verify. Writes to outside systems happen exactly once. Any step can be rolled back in one click.

That’s how implementation works: the agent builds the org chart, assigns permissions, and configures services, but you see every step as a plan card and approve it. It’s also how the personal assistant works: it handles your own items on its own, but other people’s tasks, payments, and deletions only happen after you say yes.

Sign-in and accounts

  • Two-step sign-in with a code. When a person has code verification turned on, no service will let them in with a password alone.
  • One sign-in for every service. One account across the whole system: no ten passwords to remember, and no ten accounts to close when someone leaves.

Company isolation

Every company in PRANUS is isolated from every other: employees of one company can’t see another company’s data. Conversations between companies only happen where you open them yourself, for example with guests from a partner company in the messenger.

Private stays private

  • Private conversations. A customer conversation marked Private is hidden from everyone, even admins, and never goes into the CRM.
  • Secret chats in CHATUS. On-device encryption. Organization keys are held in escrow, and we say so plainly in the interface.
  • Personal data in chat analysis. Before turning chat history into a knowledge base, the agent strips customers’ personal data.

Money without duplicate entries

In accounting, a repeated operation never creates a duplicate entry: if a request arrives twice, the money is counted once. Bank credentials, e-signature keys, and AI provider keys are stored encrypted.

Audit logs and monitoring

  • Agent change log, with rollback.
  • Log of plan changes and actions in accounting.
  • Permission log: who got which access and when.
  • Monitoring with alerts to our team and an honest health check that confirms the AI is actually answering, not just that a process is running.

Where your data lives

You choose:

  • PRANUS cloud: the storage location is specified in your contract.
  • Your own servers: data stays with you, and only one secure connection goes out, for updates and the license.
  • Air-gapped: data and AI models stay inside your perimeter, and no AI request leaves it.

Compare the options on the cloud, self-hosted, or air-gapped page.

What we don’t claim

We don’t claim certifications we don’t have, and we don’t promise what we don’t do. If your security team needs a questionnaire, a requirements review, or a call with our engineers, get in touch and we’ll answer every point.

FAQ

Can an AI agent see documents an employee isn't allowed to access?

No. Permissions are checked before the search, not after, so the model never receives documents the person isn't cleared for. The agent acts with the permissions of the person it works for, and no more.

What happens to access when someone leaves the company?

Offboarding is a single HR event in TEAMUS. It removes the person's permissions across every service, returns their tasks to the department inbox, and revokes the agents' service keys.

Can admins read employees' private conversations?

No. A conversation marked Private is hidden from everyone, including admins, and never goes into the CRM. For especially sensitive discussions, CHATUS has secret chats with on-device encryption.

Does PRANUS have ISO 27001 or SOC 2 certification?

We don't hold those certifications, and we don't claim them. If your security team needs a questionnaire filled out or specific requirements reviewed, send it to your sales manager and we'll answer every item.

Where is our data stored?

Wherever you choose: in the PRANUS cloud (the storage location is specified in your contract), on your own servers, or in an air-gapped setup together with local AI models.

Let’s bring order in two weeks

Tell us about your company — we’ll show PRANUS on your real tasks and, if we’re a fit, implement it for free.