What we track — and the numbers behind it.
AITerm publishes its own usage counters in public. Every figure on this page is an anonymous event count — no IPs, no emails, no machine names, no command history, no chat content. Read the privacy policy for the full list of fields we record (it's short).
What this page will never show
IP addresses
email addresses
usernames
machine hostnames
command history
file contents
chat / AI-prompt content
tracking pixels
third-party analytics
That's the privacy commitment. The counters below exist only to prove we operate on it — they're event types and counts, nothing more.
…
Tamper-evident audit log
Every counted event lands in a hash-chained log on disk. entry[n].prev = sha256(entry[n-1]) — flipping a single past number breaks every hash after it. The state of the chain is checked on every page load; what you see is what was logged.
…
Activity, last 7 days (the headline window — long enough to capture how the platform actually breathes)
AI backends launched — last 7 days (only what was actually started; supported set is much larger)
Snapshot right now (the very moment you loaded the page)
Today, last 24 hours (rolling — naturally noisier and can read zero during quiet hours)
How to read these numbers honestly.
Connector reconnects are heartbeats — every paired connector reconnects after a hub restart or network blip, so a high count usually means we shipped a fix, not that the fleet exploded. Logins (ok) only counts fresh sign-ins from the login form; session cookies last 24 h, so most active users don't appear here daily. Sessions running now is genuine concurrency at the moment you load the page — quiet doesn't mean unused. By-AI breakdown reflects what was started here; if only one backend shows, that's the only one in use, not the only one supported.
loading…