Straight answers. Short ones.
01Will my team feel it?
Barely. A check adds a fraction of a second to a prompt.
02What does it stop?
Instructions planted in prompts and attached files, and any agent action outside what that person’s team may touch: files, shell commands, company tools, databases and websites. Insiders may have accounts giving them legitimate access to computer systems, with this access originally having been given to them to serve in the performance of their duties; these permissions could be abused to harm the organization.
03Does it block ordinary work?
On our bench, 6 of 100 ordinary prompts were blocked, 5 of them one prompt with the word password in it. A long attached file with order-like text in a part the check could not read waits for your admin.
04What is agentic security?
Protection against agentic attacks and agentic CVEs. Kriton checks every prompt against our own defensive RAG of them, and we keep adding to it.
05What does the policy engine do?
It sets, per team, which folders, files, company tools, databases and websites its agents may touch. It runs today in the console’s Policies page.
06Which agents does it cover?
Claude Code today. We name another agent here once it is built and tested.
07Can my people switch it off?
On company-managed laptops a lock keeps it on for standard users. The lock is new, so we test it on your laptops in the pilot. A local administrator can undo any lock.
08What if Kriton is down?
Your setting decides: let short prompts through, or stop them. Agent actions keep following your rules for up to a day.
09Do you see our code?
Only what a person puts in a prompt, which goes to the model provider that runs our check. Allowed prompts are not kept, unless audit mode or an outage let them through; those and stopped ones are locked to a key only you hold.
10Can we start small?
Yes. One team and one folder. Add or remove any piece whenever you like.
11Does it replace our endpoint security?
No. It sits beside it and covers what your AI agents do.
12Can we audit it?
Yes. Your security team can audit the product anytime under NDA.