Short answer: Claude Code keeps your sessions on your machine, under ~/.claude/projects/, your skills in ~/.claude, and your memories and task notes in each project's own .claude folder. Moving to a new computer means carrying all of those, leaving your login behind, and fixing the folder paths when they change. crewrota does it in two commands: crew backup --password on the old machine and crew restore <zip> --password on the new one.

A few days ago I set up Claude Code on my laptop with the backup crewrota had then. My skills did not come over, and neither did the task notes I keep in each project's .claude folder. I keep a README and the evidence for every task there, so on a new machine I can pull the code from GitHub and carry on from the notes. Without them, the laptop was only half set up.

The problem: your setup is in more than one place

Claude Code's data usage page says where your history lives:

"Claude Code clients store session transcripts locally in plaintext under ~/.claude/projects/ for 30 days by default to enable session resumption. Adjust the period with cleanupPeriodDays."

That is one place. The others:

  • Your main setup. Skills, agents, commands and CLAUDE.md sit in ~/.claude. Some skills are links into other folders, and a plain copy may not follow them.
  • Each project's .claude folder. Tasks, memories, project skills and rules. Task folders are often kept out of git, so git clone does not bring them back.
  • Your login. It sits next to all of this. You do not want it in a zip you copy around.

And one catch: Claude Code files each session under its project's full folder path. If your home folder or drive is different on the new machine, --resume will not find the old sessions until those folders are renamed.

How to move Claude Code to a new computer

On the old machine, check what will go, then write the zip:

crew backup --dry-run crew backup --password

Copy the zip to the new machine, install crewrota there with npm install -g crewrota, and restore:

crew restore crewrota-backup-20261103-1735.zip --password --clone crew login main

The restore merges your history in, re-creates your crew members, puts every .claude folder back in its project, adds your skills to ~/.claude, and with --clone clones each repo from its remote first. Then you sign each member in again with crew login <member>. If your projects live somewhere else now, add --map <old folder>=<new folder>, for example --map E:\Projects=D:\work.

Here is the dry run on a demo setup with three projects:

[i] 3 project .claude folders and 3 skills: 23 files, 210.4 KB [i] Largest (leave one out with --skip <folder>): 87.9 KB E:\demo-crew\Projects\shop-api\.claude\tasks\checkout-bug 58.6 KB E:\demo-crew\Projects\notes-app\.claude\tasks\sync-fix 39.2 KB E:\demo-crew\Projects\shop-api\.claude\tasks\rate-limits 24.4 KB E:\demo-crew\Projects\blog-site\.claude\tasks\dark-mode 98 bytes E:\demo-crew\home\.claude [i] 1 login or key files stayed behind: E:\demo-crew\Projects\shop-api\.claude\.env [!] E:\demo-crew\Projects\notes-app: 1 unpushed commits. Only pushed code comes back on a new machine. [!] 1 files look like they hold a key (Anthropic key). Add --password to encrypt the backup: E:\demo-crew\Projects\shop-api\.claude\tasks\rate-limits\evidence\notes.txt [i] Dry run: nothing written.

What goes in, and what stays behind

You do not list anything. crew backup reads the folders Claude Code has worked in from your transcripts, then takes every .claude it finds in them.

  • Goes in: your ships' history, memory and MCP servers, the member list with each member's model and effort, each project's .claude (tasks with their evidence, memories, skills, rules), and your skills, agents, commands and CLAUDE.md from ~/.claude. Linked skills are stored as real files.
  • Stays behind: logins and key files (.credentials*, .env, *.pem, *.key and similar), settings.local.json, node_modules, .git and worktrees. Plugins are not copied: the restore lists the ones you had, with the command to install each.

The zip holds your transcripts and task notes in plain text, so use --password. It encrypts the zip with AES-256. File names inside stay readable, so do not put a secret in a file name.

If you used Claude Code before crewrota, your old sessions are in your main ~/.claude. Copy them into a ship first with crew ship import <ship> --from <folder>, pointed at that .claude folder; crewrota only reads it.

crew restore on a demo setup: the ship, the skills and three project .claude folders come back, then crew login for each member

Your accounts and Anthropic's terms

A backup never holds a login. Each member signs in again on the new machine through Claude Code's own login, and crewrota never copies or moves a login. Read Anthropic's Consumer Terms and Usage Policy for your plan.

Questions about crewrota

Can I move Claude Code to a new computer without losing my sessions?
Yes. crew backup puts your ships' session history in the zip, and crew restore renames the project folders when your home folder or drive is different, so --resume finds the old sessions.

Is my Claude login in the backup?
No. Logins and key files stay behind. After the restore, run crew login <member> for each member.

Will a restore overwrite my files?
No. A file that is the same is skipped. A file that differs arrives beside yours as <name>.incoming-1, so you can compare the two and delete one.

Does it work with file names in Sinhala, Tamil or Japanese?
Yes. Such names go into the zip as ~crew1, ~crew2 and so on, because the tar in newer Windows crashes on them, and the restore puts the real names back.

This is in crewrota 1.0.4, free and open source on npm and GitHub. If a restore misses something of yours, open an issue on GitHub.