What it reads, and what it sends

What leaves your machine

A single 1-token API call every five minutes — the rate-limit reading, which is the one thing that cannot be computed locally. That is all. There is no account to create, no telemetry endpoint, and nothing to log into: the app reuses the OAuth token Claude Code already stored, and never writes to a configuration folder.

What it reads from your transcripts

• ✓Token counts, model ids and flags — never your message content.
• ✓One visible exception: the Sessions list names each conversation, using the title Claude Code generated for it or the prompt that opened it, because a list you cannot name is a list you cannot search. Both are capped at 200 characters before they are stored, they appear on that one screen, and nothing further into a conversation is ever read.
• ✓The 24-hour insights scan is bounded to files touched in the last 24 hours and runs in the background.
• ✓Usage history is stored locally on disk and never uploaded.

What the context scan reads

Sizes, names, timestamps and frontmatter from ~/.claude — never the contents of your memories, instructions or skills — and it never edits them. The one action it offers is copying a cleanup prompt for Claude Code to act on.

What the page in Settings hides

Account holder, email and absolute paths are masked until you ask for them, because System information is the page that ends up in a screenshot. Copy for a bug report copies the page exactly as shown, so a masked holder stays masked.