What it does
LynxTrac connects your AI assistant to the endpoints you manage in LynxTrac. Ask which devices are
offline, which alerts are open and where, which machines are missing security patches, what software is
installed where, how CPU, memory and disk have trended, which services are failing, what the logs say, and
how support tickets stand, across the customers you manage that have it turned on.
It works over the Model Context Protocol (MCP), an open protocol Claude and ChatGPT support for connecting
to outside services. Your assistant asks LynxTrac for the data it needs, and LynxTrac answers with the same rules the
console applies to you.
A few questions to start with:
- Which of my servers are offline right now?
- Summarize open alerts by customer, most severe first.
- Which devices are missing security patches?
- Show CPU and memory over the last week for the busiest server.
- Where is Google Chrome installed, and on which versions?
Before you start
- A LynxTrac user account with an email address in its profile.
-
Two-factor authentication turned on for that account (My Profile in the user menu). LynxTrac asks for it
before it lets a client connect.
-
External AI clients turned on for the customers you want to ask about, with AI credits available. This is
in early access: we turn it on per customer, on request. If yours does not have it yet,
contact us.
Connecting Claude or ChatGPT
Claude lists LynxTrac in its connector directory, so you can add it there without typing anything.
ChatGPT does not list it yet, so there you add it by hand with this server address:
The menus inside Claude and ChatGPT move around between releases, so if a step below does not match what
you see, look for the place to add a connector or app by server URL.
Claude
-
Open the LynxTrac listing in Claude's connector directory (or find LynxTrac in the
directory from Settings, then Connectors, in Claude on the web or desktop) and connect it.
- A browser window opens on LynxTrac's own sign-in page. Sign in, including your second factor.
-
The Connect an AI client page shows the account the client will act as, that access is read-only, and what a
request costs. Choose Allow.
- Back in Claude, try one of the questions above.
You can also add it by hand: in Settings, then Connectors, choose Add custom connector, name it
LynxTrac and paste the server address above. Leave any client ID and secret fields empty, because
Claude registers itself. Then sign in and allow the connection as above.
The listing is just a quicker way in. Either way it is the same connection, and requests are refused until
external AI clients are turned on for your account.
ChatGPT
-
In ChatGPT, open Settings, then Apps & Connectors, then Advanced, and turn on Developer mode. On a business
workspace an admin may have to allow this first.
- Create a connector named
LynxTrac with the server address above, and choose OAuth as the authentication. - Sign in on LynxTrac's page, check the Connect an AI client page, and choose Allow.
- In a chat, turn the LynxTrac connector on and ask your question.
What it can and cannot do
A connected client can read devices and their health, alerts, patch status, installed software, services,
performance metrics, logs, support tickets and cloud inventory.
It cannot change anything. Only read operations are offered to the client at all, so there is nothing for it
to call that would restart a service, install a patch, run a script, edit a ticket or touch a setting.
The client acts as the person who connected it, with that person's permissions. Answers are limited to the
customers your LynxTrac account manages and to the features their subscription includes. If you cannot see a
device in the console, the client cannot see it either.
Credits
Each successful request uses AI credits from your account. A refused request costs nothing, and neither do
signing in or the client checking what LynxTrac offers. The Connect an AI client page tells
you how requests are charged before you allow the connection, and administrators can see credit usage in AI
Settings.
Who can see and control it
- You always sign in on LynxTrac's own sign-in page, never on a form inside the AI client.
-
Your account's administrators can see every connected client, who connected it and when it was last used,
and can disconnect any of them. Their AI settings can also switch AI off for the account.
- Every call a client makes is recorded in LynxTrac's AI audit log.
-
What LynxTrac returns becomes part of your conversation in the AI client, so it goes to the AI provider you
chose (Anthropic for Claude, OpenAI for ChatGPT) under your own account with them. Their terms and privacy
policy cover it from there. Our privacy policy explains how LynxTrac handles
your data.
Disconnecting
In LynxTrac, go to Setup > AI Settings > External AI clients (MCP) and choose
Disconnect next to the client. It loses access on its next request. To use it again, connect it
again from the client.
You can also remove the connector inside Claude or ChatGPT. Doing it in LynxTrac as well is the way to be sure
the access is gone on our side.
Troubleshooting