FAQ
| We send |
- Timestamps of your sessions
- Which tool you used (Claude Code / Claude Desktop / Codex)
- Call counts per tool
- Subagent / MCP / skill call counts and names
- Your reasoning-effort setting
- Whether a CLAUDE.md is present
- Whether plan mode was used
- Edit and verification counts
- Token counts
- A hash of your working folder name
|
| We don't send |
- Prompt text
- Tool output
- Code
- File contents
|
Do you see my prompts or code?
No, not by default. We only receive timestamps, tool name, call/token counts, and a working folder name. Data is kept 90 days and deleted immediately on cancellation.
What's the coaching based on?
Right now, on one person's own usage patterns and fixed thresholds. You're compared only to your own past week, never to other people. These thresholds are a starting point until there's data from more users.
Does this work on Windows?
Not yet. macOS is supported first; Linux has a cron template. Windows support is not available.
Do you cover ChatGPT usage?
No. This covers Claude Code, Claude Desktop, and Codex CLI/desktop usage only.
Does this work with Claude Desktop?
Yes (macOS). It reads Claude Code, Claude Desktop, and Codex logs.
What does this actually tell me — and what can't it tell me?
It can measure behavior patterns (the five practices, and 10 traps). It cannot measure the quality of your prompts themselves — that's by design, since we never send prompt text.
Can I cancel anytime?
Yes. Cancelling stops future charges and deletes your stored data immediately.