Skip to content
AI governance

Every AI tool on your fleet. Seen, mapped, governed.

Find the AI clients, coding agents, MCP servers and local models on every Mac, PC and Linux machine. See them on one map. Block what you do not allow, and know the block works.

Found

Clients, agents, MCP servers, extensions and local models, on Mac, Windows and Linux.

Mapped

Who has which tool, what runs right now, and what each one can reach.

Blocked

The app and its website together, and a plain answer when a block cannot apply.

Private

Credentials are named, never read. Prompts and conversations are never collected.

AI map

Every AI tool, by team, on one screen

Tools down the side, your groups across the top. Switch between installed, running now and used in the last week.

  • Allowed, blocked, or local only: every tool is allowed until you block it
  • New tools flagged the week they first appear
  • A blocked tool that still runs is marked, with the reason
  • Daily trends over a month, a quarter or a year, with every block marked on the line

How the AI map works

One device, traced

See what each AI tool can reach

On every device, a graph of which AI tools use which MCP servers, and what those servers can touch: files, networks, shells and credentials.

  • MCP server packages checked against the GitHub Advisory Database
  • A credential in a git repository, or a server that can run a shell, shows in red
  • Local models on their own band: nothing they do leaves the device
  • Credentials are named, never read. The value never leaves the device.
Control

Block a tool. Its website goes with it.

Blocking an AI tool stops the app launching and blocks its service, browser tabs included, on Macs, Windows PCs and Linux machines. Nothing is restricted until you choose.

  • Block a whole provider, not just one app
  • Remove an AI connector from a config file, uninstall a tool, or pin it to a version you have tested
  • Each device shows how many blocked tools are not actually stopped yet
  • An alert when a block starts, or a change fails

How AI enforcement works

Risk

A score for every device

AI risk out of 100With the findings behind it. Sending data to an outside provider counts; running a model locally does not.
Smart groups by AI postureSuch as every device with a vulnerable MCP server.
Exports and snapshotsThe inventory as JSON or CSV, and snapshots to compare over time.
Compliance

AI rules, on by default

MCP servers free of known advisoriesEvery npm and Python MCP package, checked.
No credentials in AI tool configurationPlain-text tokens found where they should not be. Location and kind only.
Off until you turn it onAI discovery changes nothing on a device, and skips the folders you exclude.
AI-ready

Ask your fleet anything

Claude, Cursor and VoxyAI can query your fleet in plain language. They can look. They can never touch.

  • Strictly read-only by construction, enforced by an automated test suite
  • Secrets like recovery keys and Wi-Fi passwords are redacted before the AI ever sees them
  • Every AI query lands in the audit log, attributed to its token

Connect an AI assistant

AI assistant - connected to Xavier
Any Windows laptops with BitLocker off?
Two of your 14 Windows devices report BitLocker disabled: WIN-4F2KQ in Sales and WIN-9QLMB at the front desk. Both are marked non-compliant against the BitLocker rule.
Which devices haven't checked in for two weeks?
Three devices are past the two-week mark: a MacBook Air assigned to a departed contractor, an iPad in the conference room, and one Android tablet. Want the full inventory details for each?

See the AI on a fleet that is already there.

The live demo has the AI map, every device's AI graph and the AI report, on 150 sample devices.

Try the live demo Start a 30-day trial