Missing from the sidebar is not missing from the disk.

Every conversation is a .jsonl transcript under ~/.claude/projects, in a folder named after the working directory with every non-alphanumeric character replaced by a dash. Claude Desktop keeps a separate list of those sessions as local_<uuid>.json files under claude-code-sessions. Two different sets of files, and nothing reconciles them.

Usually just missing from the list. The sidebar is built from index records while the conversations live in separate transcript files, so when a record is lost the session vanishes from the sidebar with the transcript untouched on disk. Open ~/.claude/projects and look at the file sizes to check — you do not need this app to find that out.

Claude Desktop keeps a separate index per account, under claude-code-sessions/<account>/<organization>/. Signing into a second account gives you a second, empty list. Your transcripts never moved and were never account-specific; only the list of them is.

The index record holds the title and a pointer called cliSessionId naming the transcript file. When that pointer is missing the sidebar still draws the row from its cached title, but there is nothing to load. Scan & Repair searches the transcripts by title to find the file again.

No. There is no account, no sign-in and no telemetry. The only network request the app makes is a version check against GitHub, which can be switched off. Everything else happens on your disk.

Every write shows a preview first that cannot be skipped, anything replaced is backed up before the write, and every change is appended to a plain-text audit log. Copying between accounts only adds. The one control that removes anything is called Remove, it unlinks a session from one account's list, and it never touches the transcript.

Reasonable question. Start with Export: scanning and exporting only read, so you get a zip of your history without a byte written to Claude's files. If you never press anything else, you have still gained a backup.

Not yet. The code paths for both are implemented and unit-tested — the app knows where Claude keeps its files on each platform — but no build has ever been produced or launched on either. Only the Windows build is real today.

Export writes every session you select into a single zip, carrying the original bytes of each transcript plus a checksum for each one. Import reads it back on another machine. Export only reads, so it is the safe thing to do first — and doing it before Claude Code's retention period removes anything is the only defence against the case nothing can fix.

A .jsonl transcript on disk that no account lists in its index. Sessions started from the CLI or the VS Code extension often never get a record written, so they never appear in the Claude Desktop Code tab. This app reads every transcript regardless of who listed it, so orphans show up, can be checked for damage, and can be exported.

No, and that is deliberate. It repairs desktop history records that lost their pointer to a transcript, which is a different and much narrower thing. It does not write to Claude Code's own resume index, so it will not change what /resume or claude --continue offer you.

Export on the old machine, copy the zip across, Import on the new one. Import shows you exactly what it would write before it writes it, refuses anything whose checksum does not match, and never overwrites an existing transcript without backing the old one up first.