Reaching a human
We are a small team, which means no ticket-deflection maze and no tier-one script, but also no 24/7 phone line. Both halves of that are worth being honest about.
- Email support and you get a reply within one business day. Include the
x-lcllm-request-idand you will usually get an answer rather than a follow-up question. - Live incidents go to the status page first. Check it before writing in, because it is faster than we are.
- Dedicated-tier customers get a shared Slack or Discord channel and a named contact.
Uptime and credits
Our target is 99.9% monthly availability, measured as successfully proxied requests over total well-formed requests. Failures caused by an upstream provider outage are excluded, because we cannot be more available than the API we forward to, and pretending otherwise would make the number meaningless.
| Monthly availability | Service credit |
|---|---|
| < 99.9% | 10% of that month's spend |
| < 99.0% | 25% of that month's spend |
| < 95.0% | 50% of that month's spend |
Credits are applied to your balance on request, within one billing cycle. You have to ask. We would rather say that plainly than imply an automatic process we do not run yet.
Incident process
- Detection to status page: under 10 minutes. Automated, not dependent on someone being awake.
- Updates every 30 minutes while an incident is open, even when the update is “still investigating”.
- Public write-up within five business days for anything that breached the availability target, including what we got wrong.
Security issues do not go through support. Email security directly. That inbox is monitored on a different schedule.