How do I set up AI usage telemetry?
Setting telemetry up is a two-part job: create a token in Velory, then push the configuration it generates to your team's machines through your MDM. You do it once for the whole fleet rather than per service.
The configuration is generated for you. You are distributing files, not writing them.
In this guide
Before you begin
You need Admin access, and the API Tokens page requires full platform permissions.
You also need a way to push files and environment variables to your team's machines — an MDM, a configuration-management tool, or an onboarding script. Velory generates the configuration; getting it onto laptops is your side.
Create the telemetry token
- Go to Settings → API Tokens.
- Click Create token.
- Give it a name you will recognise later — the name is only ever shown on this page and is never sent with a request.
- Under What is this token for?, choose AI usage telemetry.
- Set Expires in.
- Click Create token.
Choosing the telemetry purpose matters. That token can submit AI-tool usage data and nothing else — no API, no MCP, no read access to your data. If it leaks, the worst somebody can do is send you usage records. Do not reuse an API access token here; it is not allowed to submit telemetry anyway.
Download the configs
Velory shows the token once, on creation, alongside a Telemetry configs panel. This is the only time you will see either.
Download all three before closing the dialog:
| Tool | What you get |
|---|---|
| Claude Code | A config file with the token and endpoint embedded |
| Codex | A config file with the token and endpoint embedded |
| Gemini | An environment variable — Gemini takes no config header |
Then copy the token itself and confirm you have it. If you close the dialog without downloading, the token cannot be retrieved: revoke it and create another.
Push them to your fleet
Distribute through whatever manages your machines:
- Claude Code and Codex — place each config file where the tool expects it.
- Gemini — set the environment variable shown in the dialog, so it is present for the shell your team runs Gemini from.
The files embed both the token and the ingest endpoint, so there is nothing further to configure on the machine.
Reporting starts the next time somebody uses the tool. Usage is recorded per day, so allow a day before judging whether it is working.
Check that it is arriving
Open the matching service in Assets → Software and go to its Usage tab, then the Telemetry sub-tab. Rows appear as usage lands. Until then the tab says "No coding-tool telemetry yet".
Licence health scores take longer. They are recalculated once a day, so expect a grade to appear the day after usage first arrives, not the same afternoon.
Good to know
Nothing has appeared. What do I check?
In order:
- Has anyone used the tool since the config went out? Nothing is reported retroactively.
- Did the config actually land? Check one machine by hand before suspecting Velory.
- Does the service exist in your catalogue under the name the tool reports as — Claude, OpenAI or Gemini? Usage has nowhere to attach otherwise.
- Has the token expired or been revoked? The API Tokens page shows both.
- Give it a day. Usage is aggregated daily.
The token expired. What breaks?
Reporting stops silently — the tools keep working, they simply stop being heard. Tokens cannot be extended, so create a new one, distribute the new configs, and revoke the old token.
Set a reminder before the expiry date. There is no warning when a telemetry token lapses, and the first symptom is health scores quietly going stale.
Can I use one token for several companies?
Telemetry tokens are scoped to the company that created them. An organisation with several companies needs one token and one rollout per company, so usage is attributed to the right place.
Someone shows as "Unmatched". Why?
The email in the tool's telemetry does not match any employee record. Correct the address on the employee — matching is retried as new usage arrives, so it resolves itself once the two agree.
Do I need to do anything per service?
No. One token and one rollout covers all three tools. Adding a Claude plan later needs no telemetry work; usage attaches to the service by name.
Was this article helpful?