Skip to content
Menu

SOLUTIONS / START WITH THE QUESTION

What do you need help with?

Start with a situation that sounds familiar. Each fictional example shows a question to ask, the evidence to inspect and a decision a person keeps. The examples do not run the software. The products are in development, and Stillpoint is paused.

01 / AGENT DEFENDER

Check permission
before the action.

You ask an AI tool to prepare a report for your team. It then requests permission to send the report outside the team. That is a new action with a new recipient. Agent Defender is designed to compare the request with the task’s permissions before a connected action proceeds.

In development
Explore Agent Defender

Only tools and actions set up to use the permission check can be covered.

ILLUSTRATIVE SCENARIONOT A TEST RESULT

What is it asking to do?
What was allowed?

Proposed action
Send an internal report to an external destination
Permission
Prepare a report for internal review
Question
Does that permission include this destination?
Why does the decision change?

Permission to prepare the report does not include permission to send it outside the team.

A useful permission rule names the action and its destination. The intended check compares those details with the permission given, then records the decision and reason.

An example only. No real request was checked, allowed or stopped here.

02 / SWARM HUNTER

Bring the events together.
See what they tell you.

A fictional website records failed logins followed by requests for private information. Seeing the events together gives a reviewer a sequence to check. Swarm Hunter supports that review with the records behind each concern.

In development · Local review
Explore Swarm Hunter

Connections to live systems still need testing. A person reviews the findings; activity is not automatically blocked.

FICTIONAL ACTIVITY RECORDSEXAMPLE ONLY

A sequence to investigate.

  1. Login attempts refused

    Several sessions show failed logins.

  2. Requests for sensitive records

    Requests reach the same fictional service.

  3. REVIEW
    Check ordinary explanations

    Check for an expected automated task, an account problem or missing records before concluding that the sequence is hostile.

What can’t these records tell us?

These records do not prove who caused the activity, whether AI was involved or whether harm occurred.

A person would check the records, the access that was logged and any missing information before acting.

This example was created for explanation. It does not show a live incident or a working detector.

03 / MALWARE SCOUT

A closer look.
An informed next step.

Activity on a Windows computer has raised a concern. Malware Scout’s research version looks for patterns in records you supply. A repeated pattern can give a person something specific to examine; it does not establish that the computer is infected.

Experimental · In development
Explore Malware Scout

Not a replacement for antivirus. No public download or active scan is offered here.

DISCUSSION FRAMEWORKNOT A SCAN

Start with what happened.

What changed?
Describe the behavior that raised the concern.
What is known?
Record what you observed, when it happened and what remains a suspicion.
What comes next?
Identify the records that could confirm an explanation or show that more review is needed.
What doesn’t Malware Scout do?

Scanning, monitoring and removal still need development and testing. Installing software does not prove it protects the computer.

Do not submit suspicious files, live malware, credentials or confidential records through the contact page.

04 / STILLPOINT

A planned workspace.
Development paused.

A planned place to prepare a draft, calculation or proposed change on your own computer, then review the details before acting. Development is paused; it is not an available service.

Read the Stillpoint status
Development paused

Stillpoint isn’t available yet.

There is no service or customer connection to activate. Privacy and security have not been independently proven.

The preview uses fictional results rather than taking real actions.

Tell us about the problem.

Describe what happened or what you want AI to do, and the result you need. We can discuss whether a focused test would answer that question. Keep private records and passwords out of the message.

Discuss your needs