AxLoop

How it works

Observe the interaction from the edge outward.

AI-mediated work can begin on a laptop, in an IDE, or inside a third-party assistant before a request reaches software you control. AxLoop starts with supported evidence at that origin.

Discover what exists. Map what connects. Build toward traceable outcomes.

Request a demo See the architecture

How the platform evolves

Evidence first. Control later.

AxLoop does not ask customers to write policy against an interaction it cannot yet explain. Discovery establishes the inventory. Correlation maps relationships. Later stages can trace, evaluate, govern, enforce, and audit supported interactions.

  1. 01available now

    Discover

    What AI exists?

  2. 02near term

    Map

    What is connected?

  3. 03roadmap

    Trace

    What happened?

  4. 04roadmap

    Understand

    Did it work?

  5. 05roadmap

    Govern

    Should it have happened?

  6. 06roadmap

    Enforce

    What should happen next?

Available now · Discovery pipeline

One continuous flow, on the device.

01DeviceApps, agents, runtimes, extensions, connections
02SanitizeSensitive fields removed or rejected
03ReconcileCompared with previous state
04QueueOnly changes enter the durable outbox
05ExportOpenTelemetry to your collectoroptional
06AnalyzeFleet-wide AI inventory

If nothing changed, nothing is exported. If something changed, only the new state leaves the device — less telemetry, more signal.

Evidence, not assumptions

Installed is not running. Configured is not connected.

AxLoop reports what the endpoint actually proved — never more than the evidence supports.

Installed

A supported app or binary exists on the device.

Configured

A configuration declares a server or integration.

Running

A supported process is present right now.

Observed

AxLoop has direct evidence for the state reported.

Privacy at the data model

What AI exists — not what anyone typed into it.

Sensitive values are filtered before telemetry is created. Known secret patterns are rejected, not exported. Paths are reduced to safe identifiers or hashed locally.

Local components talk over authenticated Unix sockets. No open ports. Export is off until you turn it on.

  • Prompts or model responses
  • Source code or file contents
  • Command-line arguments
  • Environment variables
  • URLs or auth headers
  • Credentials or API keys
  • Configuration contents
  • Full home-directory paths

Two scopes

The whole machine, and the person using it.

This Mac

The system service observes host-level signals — supported running processes and machine-level assets.

My Session

The user-session agent discovers user-specific tools and configuration — editor extensions and declared connections stay in the user scope.

Enterprise deployment

Deploy the same way you deploy endpoint software.

A native .pkg built for device management — daemon, session agent, menu-bar app, and local inventory in one package.

Jamf
Kandji
Microsoft Intune

Lightweight by design

  • Native Rust + Swift
  • No Python runtime
  • No local web server
  • No open TCP port
  • Bounded filesystem reads
  • Battery-aware scans
  • Differential telemetry
  • No GPU workload

AxLoop AI

Discover the AI edge you already have.

See AxLoop on a real device.

Request a demo Explore device coverage