VISION

Understand AI.
Make informed decisions.

My vision for Nyxus is practical: give organisations the context to decide how AI belongs in their work, while keeping control of their own data.

“Knowing a tool is installed is a starting point. I want the platform to help people understand the use behind it, and decide what happens next.”
Aman KarirAman KarirCreator, Shadow AI Guard & Nyxus
THE DIRECTION

Visibility should lead to understanding.

01 · CONTEXT

Look beyond the tool name.

The account, the device and the way a tool is used all matter. My aim is to bring that context together so an observation becomes something people can act on.

02 · JUDGEMENT

Keep people in the decision.

Governance needs an owner, a reason and a way to review a decision. The platform should support that judgement and make it easier to follow through.

03 · CONTROL

Keep the data close.

Self-hosting is central to the approach. Organisations should be able to investigate their AI use within infrastructure they control.

04 · PRACTICE

Make it useful day to day.

My direction for Nyxus is a practical enterprise workspace: a place to understand findings, record decisions and return to them as circumstances change.

WHERE IT STARTED

One question.
Seven stages.

What AI is actually running at work? I started with a weekly script to answer that question. It grew into Shadow AI Guard, and then into the foundation for Nyxus. The scope has grown; the reason for building it remains the same.

  1. 01 / 07

    A Python script

    I started with a Python script that checked our existing tools for known AI services and printed a report.

    It worked, but only when I ran it. By the time I read the results, things had often moved on.

  2. 02 / 07

    Running on our existing stack

    I moved the scanners onto our Kubernetes cluster and scheduled them to run automatically. Findings went into Grafana, with warnings sent to Slack.

    Then I added a browser extension and endpoint collector to see which accounts people were using across browser, CLI, IDE and desktop AI tools.

  3. 03 / 07

    Bringing Linux into view

    Our Linux machines sat outside MDM, leaving a gap. I brought them into scope through our RMM tools and other routes that didn’t depend on MDM.

  4. 04 / 07

    Paste Guard and browser security

    I wanted to do more than spot AI use. I added Paste Guard and browser AI security, with their findings going into the same place as everything else.

  5. 05 / 07

    Shadow AI Guard

    Grafana helped me see what was happening. I still needed somewhere to follow it up.

    Shadow AI Guard brought the findings into one console, with a place to review them and decide what to do.

  6. 06 / 07

    Making it easier to run

    Next, I packaged Shadow AI Guard so other teams could run it too. The managed version supports Docker Compose and Kubernetes, with one command to get started.

  7. 07 / 07

    Nyxus

    Nyxus builds on that foundation as an enterprise workspace. It’s the next stage of the same idea: understand what’s being used, then decide what happens next.

    It’s coming soon.

STILL BUILDING

1,182 commits
since 21 July.

Shadow AI Guard is open source; Nyxus is being built privately alongside it. This is both repositories together, a day at a time, and every week since the first commit has some.

1,182 commits since 21 Jul 2026
Commits
1,182commits
since 21 Jul 2026
Busiest day
77commits
22 Sept 2026
Longest streak
29days
21 Aug – 18 Sept
Days with commits
83of 92
every week has some
Hover a day for its commits · arrow keys to explore
FewerMore

Snapshot of commit dates on 20 October 2026, from the public shadow-ai-guard repository and the private Nyxus repository. Automated commits are left out.Shadow AI Guard on GitHub ↗