Start free
One full audit of your logs, free. No card, no install, nothing in your request path.
We read counts and metadata only — model, timestamps, token totals. Prompt and completion text is never stored.
Don't have an export yet?
Get your logs
Five ways in. All of them send counts, never content.
Connect a provider — Fastest — 2 minutes
We read your usage reports directly from OpenAI or Anthropic with a read-only key. Daily, automatic, nothing to export.
What we receive: We receive token counts per day and model. Your prompts never leave your provider — the usage API does not contain them.
Claude Code / agent transcripts — One command
Run our exporter over your local transcript folder. It reads the counts out of the session files and writes a JSONL you upload here.
What we receive: The exporter reads token counts and timestamps only. It never copies a prompt or a completion into the export — you can read the file before you send it.
OpenAI request logs — Export + upload
Export your completions usage from the OpenAI dashboard, or ship your own request log in JSONL.
What we receive: We keep the counts and drop everything else at the door, including any text your export happens to contain.
Anthropic request logs — Export + upload
Export usage from the Anthropic console, or upload your own request log.
What we receive: Same at the door: counts in, content dropped, nothing about your prompts stored.
Anything else (CSV) — Bring your own
Any per-call CSV with a timestamp, model, and token counts works. Column names are flexible; we tell you what we matched.
What we receive: You choose the columns you send. If your file has a text column we ignore it and never store it.
or with your email
We'll email you a sign-in link. No password to remember — it works once and expires in 15 minutes.
- One full audit free — no card, ever.
- We read counts and metadata only, never your prompts.
- Logs are analyzed then deleted; nothing retained beyond 7 days.
Already have an account? Log in.