Quick answer
Anonymous usage stats are aggregated counters, such as daily token totals or feature counts, sent without identifiers. They are anonymous in the sense that no single record points to you, but small groups and rare events can still be identifying. Read what is collected, not just the word anonymous.
What 'anonymous' can mean
| Layer | Contains | Risk |
|---|---|---|
| Counters | Totals per day | Low |
| Cohorts | Group totals | Low if groups are large |
| Pseudo-IDs | Stable random ID | Linkable over time |
| Events | Feature names and timing | Can fingerprint usage |
| Content | Prompts or code | High, never anonymous |
What to check before enabling
- Is the setting opt-in or on by default?
- What exactly is sent: counters, IDs, or content?
- Is there a stable identifier that links sessions?
- How long is it retained?
- Can you turn it off later and delete prior data?
A tracker that reads local files does not need telemetry at all. That is the simplest privacy answer: nothing is sent because nothing needs to be.
Local-first as the default
UsageAtlas works locally without an account. If you enable optional sync, only aggregates are needed; prompt content is never part of usage accounting. That keeps the privacy question separate from the cost question.
The local vs provider data guide explains which sources exist, and the token types guide defines the terms.
Questions, answered
Are anonymous usage statistics really anonymous?
They are aggregated rather than identifying, but small cohorts and stable IDs can make them linkable. The label alone is not a guarantee.
Does UsageAtlas send usage statistics?
Local tracking sends nothing. Optional cloud sync is separate and documented, and prompt content is not required.
Should I disable telemetry in my coding tools?
It does not change usage accounting or cost. Decide based on what the vendor collects, not on whether you want to save tokens.