What data leaves your machine
Every request Clauder makes, the exact fields it sends, and what never leaves your computer.
Without sync, Clauder sends nothing to clauder.fyi. The extension only asks Anthropic for your usage, using Claude Code's existing sign-in. With sync on, it sends token counts, model names, timestamps and limit percentages: never prompts, commands or file paths.
This page lists each request, the fields in it, what stays on your computer, and how to stop sending data. The privacy policy covers retention, deletion and your rights.
Does Clauder need my Anthropic API key?
No. Clauder never asks for an Anthropic API key. It reads the sign-in token Claude Code already saved, from the system credential store or ~/.claude/.credentials.json, and uses it for one request:
| Request | Sent to | Contains |
|---|---|---|
GET /api/oauth/usage | api.anthropic.com, directly from your machine | Your Claude Code sign-in token in the Authorization header |
Anthropic answers with your 5-hour and weekly percentages and reset times. The token is never sent to clauder.fyi.
What stays on my machine?
- Prompt text and model responses
- Bash command text
- Working directories, file paths and file contents
- Your Claude Code sign-in token
The extension reads Claude Code's session logs in ~/.claude/projects and keeps only the timestamp, token counts, model name and request id of each entry. The project directory becomes a one-way hash on your machine: the first 16 hex characters of its SHA-256.
The statusline installer adds hooks that log tool calls and prompts to ~/.clauder/hook-cache.jsonl. That file holds up to 2,000 characters of each prompt, up to 500 characters of each Bash command, and the working directory. The extension reads only the tool name, timestamp and session id from it, and removes the lines it has synced.
What does dashboard sync send to Clauder?
Sync runs only when clauder.sync.enabled is on and a license key is set. Every 30 seconds by default, the extension sends one request to POST https://clauder.fyi/api/sync with these fields:
| Field | Value |
|---|---|
license_key | Your Clauder license key |
timestamp | When the sync ran |
current_5h_utilization_pct | 5-hour usage percentage from Anthropic |
current_weekly_utilization_pct | Weekly usage percentage from Anthropic |
last_interaction_at | Last time your usage went up |
sessions[] | Per request: timestamp, input tokens, output tokens, model name, project hash, request id |
tool_usage[] | Per tool call: timestamp, tool name (such as Bash or Edit), Claude Code session id |
After each sync the extension asks GET /api/predict for a prediction, passing the license key. The server keeps synced rows to build your history and predictions.
What does the terminal statusline send?
| Request | When | Contains |
|---|---|---|
| api.anthropic.com usage endpoint | When cached usage is older than the cache time (30 seconds by default) | Claude Code sign-in token from the macOS Keychain |
clauder.fyi/api/predict | Only with a license key in ~/.claude/clauder.conf: for the prediction element and to check whether a trial has ended | Your license key |
clauder.fyi/api/scripts/version | Every 4 hours | Nothing; it downloads updated scripts when a new version exists |
How do I stop sending data to Clauder?
- Turn off
clauder.sync.enabledor clearclauder.sync.licenseKey. The extension needs both to sync. - Remove
LICENSE_KEYfrom~/.claude/clauder.confto stop statusline prediction requests. - Delete
~/.clauder/hook-cache.jsonland the tracker hooks in~/.claude/settings.jsonif you do not want prompts logged locally.
The usage readout in the status bar keeps working without sync, because it talks only to Anthropic. Questions go to info@hellobussin.com.