Skip to main content

Portable AI coding session transfer between machines

Project description

🚂 commuter

Take your AI coding sessions to work and back.

Commuter transfers Claude Code sessions between machines. Start a session on your home desktop, export it, commute, import it on your office laptop — and pick up exactly where you left off, with full local file access.

No cloud dependency. No VPN. No SSH tunnels. Just a JSON file in your Dropbox.

pipx install commuter

commuter demo

Why?

Claude Code's Remote Control lets you view a session from your phone — but the session runs on the original machine. If that machine sleeps, loses Wi-Fi, or you need local file access on a different computer, you're out of luck.

Commuter solves this by migrating the session itself — conversation history, project config, everything — so the AI picks up right where you left off on the new machine.

The daily workflow

Morning at home:

cd ~/projects/my-app
claude
# ... work for an hour ...

# Time to leave
commuter export --latest -o ~/Dropbox/session.json

   Exported session a1b2c3d (47 messages, 12KB)
   Git snapshot: branch feature/auth @ a1b2c3d
   Saved to ~/Dropbox/session.json

Arrive at office:

commuter import ~/Dropbox/session.json

   Mapped to local path: /Users/you/projects/my-app
   Git check: feature/auth @ a1b2c3d    Restored conversation (47 messages)

  Launching Claude Code with restored session...

Claude remembers everything. You continue:

You: Let's continue with the auth module. Where were we?
Claude: We were adding OAuth2 support to the login endpoint...

End of day — back home:

commuter export --latest -o ~/Dropbox/session.json
# commute home...
commuter import ~/Dropbox/session.json

   Session continuity detected: 47  112 messages
   Replacing local session with imported version
  Launching Claude Code with restored session...

Round-trips are seamless. Commuter detects that the imported session is a continuation of the local one and replaces it automatically — no prompts, no flags.

Setup

Install

pipx install commuter

Or if you prefer pip: pip install commuter

Path mapping (if your machines have different paths)

# Tell commuter how paths map between your machines
commuter config set path-map "/home/you/projects" "/Users/you/projects"

If both machines use the same paths, skip this step.

Prerequisites

  • Python 3.10+
  • pipx (recommended) or pip
  • Claude Code installed on both machines
  • A shared filesystem between machines (Git, Dropbox, Syncthing, Google Drive, USB stick — anything that moves a file from A to B)

Commands

Command Description
commuter list Show all Claude Code sessions on this machine
commuter export <id> -o file.json Export a session to a portable bundle
commuter export --latest -o file.json Export the most recent session
commuter import file.json Import a session and launch Claude Code
commuter config set path-map "A" "B" Set up path translation between machines
commuter push Export current directory's session to transfer dir
commuter pull Import all pending sessions from transfer dir

Import flags

Flag Description
--project-dir PATH Override auto-detected project directory
--replace Force-replace existing session without prompting
--no-launch Import without launching Claude Code
--dry-run Preview what would happen

Push / pull shortcut

For an even faster workflow, configure a shared transfer directory once:

commuter config set transfer-dir ~/Dropbox/.commuter/

Then, from each project you want to transfer:

cd ~/projects/my-app  &&  commuter push
cd ~/projects/other   &&  commuter push

push exports the session for the current directory. Run it once per project you're taking with you.

On the other machine, a single pull picks up everything:

commuter pull

   Restored conversation (47 messages)    # my-app
   Restored conversation (31 messages)    # other

  Imported 2 session(s). To resume:
    cd ~/projects/my-app  && claude --continue
    cd ~/projects/other   && claude --continue

How it works

Commuter exports a session as a single JSON bundle containing:

  • Full conversation history (messages, tool calls, results)
  • Project directory path
  • Project config (.claude/settings.json, CLAUDE.md)
  • Git state snapshot (branch, commit, dirty files)
  • Environment metadata

On import, it restores the session into Claude Code's local storage, translates paths if needed, validates the git state, and launches Claude Code with the restored conversation.

Session continuity

Commuter tracks session lineage by hashing the first N messages. When you import a session that's a continuation of one already on this machine (same beginning, more messages), it replaces automatically. Divergent sessions prompt for confirmation.

This makes the home → office → home round-trip work without friction.

Architecture

commuter/
└── backends/
    └── claude_code.py    # Claude Code backend
    └── (future: codex, gemini, etc.)

The Claude Code-specific logic is isolated behind a backend interface. Adding support for other AI coding tools (Codex CLI, Gemini CLI) is a matter of writing a new backend — the CLI and bundle format stay the same.

Requirements

Commuter assumes your project files are already synced between machines. It transfers the session (conversation + config), not the codebase. Use Git, Dropbox, Syncthing, or whatever you normally use to keep your code in sync.

FAQ

Does my home machine need to stay on? No. That's the whole point. Unlike Remote Control, commuter migrates the session entirely to the new machine. Your home desktop can sleep, shut down, or catch fire (please don't) — the session continues independently on the target machine.

What if I forget to commit before leaving? Commuter warns you. It snapshots the git state on export and checks it on import. If there are dirty files that aren't present on the target machine, you'll see a warning. The session still imports — you just might need to sync those files.

What if my git branches are different on the two machines? Commuter warns you and asks for confirmation. Claude will have context from the original branch, which might be confusing if you're now on a different one.

Can I use this without Dropbox? Yes. Dropbox is just an example. You can use any method to move the JSON file between machines: a USB drive, scp, email it to yourself, AirDrop, carrier pigeon with a flash drive. Commuter doesn't care how the file gets there.

Does it work with Codex CLI / Gemini CLI / Cursor? Not yet. The architecture supports multiple backends, but only Claude Code is implemented today. PRs welcome.

Contributing

Contributions welcome. The most impactful areas:

  • New backends — add support for Codex CLI, Gemini CLI, or other tools
  • Bug reports — especially around edge cases in session discovery and import
  • Platform testing — tested on Linux ↔ macOS; Windows support is untested

Credits

Conceived and directed by Ljubomir Buturovic. Built with Claude Code. Jokes by Claude

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

commuter-0.1.4.tar.gz (1.9 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

commuter-0.1.4-py3-none-any.whl (20.0 kB view details)

Uploaded Python 3

File details

Details for the file commuter-0.1.4.tar.gz.

File metadata

  • Download URL: commuter-0.1.4.tar.gz
  • Upload date:
  • Size: 1.9 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.11

File hashes

Hashes for commuter-0.1.4.tar.gz
Algorithm Hash digest
SHA256 54da62debfa3f7aa6940423997a68e12fb692df9bd08a56bff61541d24a7a189
MD5 d3f887bb9b60416ef4e77443983336e2
BLAKE2b-256 4d77f2d222771d0969f38fa540329cc0748580b4340adb9cd484ac8097ee9a36

See more details on using hashes here.

File details

Details for the file commuter-0.1.4-py3-none-any.whl.

File metadata

  • Download URL: commuter-0.1.4-py3-none-any.whl
  • Upload date:
  • Size: 20.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.11

File hashes

Hashes for commuter-0.1.4-py3-none-any.whl
Algorithm Hash digest
SHA256 754f0cf1267d932be19f19af7893bed0b7c10cfcf8c3b80f95b930dc11a08232
MD5 07b4f610a94b28622222c2eafbfc4de9
BLAKE2b-256 921e03496de89e08732b44247784fe75026158eb5e069a2c70508c62d888bf71

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page