Cloud-Syncing Your Claude Workspace Across Macs with iCloud: A Clever DevOps Hack

By the ClawList.io Team | DevOps | AI Automation

---

If you've ever worked across multiple machines — say, a MacBook for on-the-go sessions and a Mac Mini as your desk powerhouse — you know the pain of keeping your development environment in sync. Dotfiles, configs, project states, custom scripts: the list of things that drift out of alignment is endless. Now add an AI agent like Claude into the mix, and suddenly you have another layer of complexity to manage.

That's exactly the problem developer @dontbesilent solved with a beautifully simple, slightly unconventional approach: put Claude's working directory inside iCloud Drive. The result? A seamlessly cloud-synced AI agent workspace that travels automatically between devices — no Git commits, no rsync scripts, no third-party sync tools required.

In this post, we'll break down how this trick works, why it's surprisingly powerful for AI automation workflows, and how you can replicate it for your own Claude-powered setup.

---

What Is Claude's Working Directory and Why Does It Matter?

When you run Claude Code (Anthropic's agentic CLI tool) or build OpenClaw skills, Claude operates within a working directory — a folder on your filesystem that serves as the root context for the agent. This directory typically contains:

• CLAUDE.md — your persistent instructions, project context, and behavioral rules for Claude
• Skill definitions — reusable task templates and automation scripts
• Project files — code, configs, data the agent reads and modifies
• Memory artifacts — notes, logs, and outputs Claude generates across sessions

Think of this directory as the "brain" of your Claude agent. It's what gives Claude continuity between sessions and what makes your custom workflows actually yours. Lose sync between machines, and you're essentially running two different agents with two different personalities and two different skill sets.

This is why the working directory is the most critical piece of state to keep synchronized when you're operating across multiple devices.

---

The iCloud Trick: Simple Setup, Surprising Power

The core idea is disarmingly straightforward. Instead of storing your Claude workspace in a standard local directory like ~/projects/claude-workspace, you place it inside your iCloud Drive folder:

# Move your existing Claude workspace into iCloud Drive
mv ~/projects/claude-workspace ~/Library/Mobile\ Documents/com~apple~CloudDocs/claude-workspace

# Or create a fresh workspace directly in iCloud
mkdir -p ~/Library/Mobile\ Documents/com~apple~CloudDocs/claude-workspace
cd ~/Library/Mobile\ Documents/com~apple~CloudDocs/claude-workspace

For convenience, you'll probably want a symlink so you don't have to type that path every time:

# Create a friendly symlink in your home directory
ln -s ~/Library/Mobile\ Documents/com~apple~CloudDocs/claude-workspace ~/claude-workspace

# Now you can just do:
cd ~/claude-workspace

Once iCloud Drive is enabled on both your MacBook and Mac Mini, Apple's sync engine takes over automatically. Any file you create, modify, or delete in that directory gets pushed to iCloud and pulled down on your other machine — usually within seconds on a fast connection.

Your CLAUDE.md stays consistent. Your custom OpenClaw skills are available everywhere. Your project context follows you from device to device without you lifting a finger.

What Gets Synced Automatically

With this setup, here's what travels seamlessly between your Macs:

• CLAUDE.md files — your agent instructions and persistent context
• Custom skill scripts — Python, Shell, or JavaScript-based OpenClaw automations
• Project scaffolding templates — starter files and boilerplate you've taught Claude to use
• Agent output logs — useful for reviewing what Claude did in a previous session
• Configuration files — API keys references (though store actual secrets in a keychain, not plaintext!)

~/claude-workspace/
├── CLAUDE.md                  # Global agent instructions
├── skills/
│   ├── code-review.md         # Custom OpenClaw skill
│   ├── blog-writer.md         # Another reusable skill
│   └── data-pipeline.md       # Data automation skill
├── projects/
│   ├── project-alpha/
│   │   ├── CLAUDE.md          # Project-specific instructions
│   │   └── ...
│   └── project-beta/
└── logs/
    └── session-history/

---

Practical Use Cases for Multi-Mac Claude Workflows

This setup unlocks some genuinely useful real-world workflows. Here are a few scenarios where cloud-synced Claude workspaces shine:

1. Laptop-to-Desktop Handoffs

Start a coding task on your MacBook while commuting or in a coffee shop. Claude builds context, generates files, and updates your CLAUDE.md with progress notes. When you get home and sit down at your Mac Mini, open the same workspace and Claude picks up exactly where it left off — same context, same skills, zero friction.

2. Shared Skill Libraries Across Projects

Maintain a skills/ subdirectory in your iCloud-synced workspace with reusable OpenClaw skill definitions. Whether you're starting a new project on your laptop or your desktop, your entire library of custom automations is immediately available without copying files manually.

3. Persistent AI Agent Memory

Use your synced workspace as a lightweight agent memory layer. Configure Claude to write session summaries or task logs to files in the workspace. Over time, you build a searchable history of what your AI agent has done — accessible from any Mac you own.

# Example: Session log entry Claude writes to ~/claude-workspace/logs/2025-07-15.md

## Session Summary - July 15, 2025
- Refactored authentication module in project-alpha
- Created 3 new unit tests for UserService
- Flagged potential security issue in token refresh logic
- Next steps: Review PR #47, update API documentation

4. Team Collaboration (with Caveats)

If your team all uses Macs and shares an iCloud Family or organizational account, you could theoretically share a workspace folder. This is more experimental territory, but for small teams or pair-programming scenarios, it opens interesting possibilities for collaborative AI agent workflows.

---

Things to Watch Out For

Like any creative workaround, this approach has a few rough edges worth knowing about:

• iCloud sync conflicts: If you're editing files on both machines simultaneously, iCloud may create duplicate conflict files. Treat your workspace like a "one active machine at a time" environment to avoid this.
• Sync delays on large files: iCloud can be slow to sync large binary files or directories with thousands of small files. Keep your workspace lean.
• Security considerations: Never store plaintext API keys, tokens, or credentials in your iCloud-synced workspace. Use macOS Keychain or environment variables for sensitive values instead.
• .gitignore discipline: If your workspace is also a Git repo, make sure your .gitignore is clean — you don't want to accidentally commit iCloud-specific metadata files like .icloud stub files.

# Add to your .gitignore if workspace is also a git repo
echo "*.icloud" >> ~/claude-workspace/.gitignore
echo ".DS_Store" >> ~/claude-workspace/.gitignore

---

Conclusion: Elegant Solutions Don't Have to Be Complex

There's something deeply satisfying about solving a complex problem with the simplest possible tool. @dontbesilent's iCloud workspace trick is a perfect example of developer pragmatism: instead of building a custom sync infrastructure or reaching for yet another SaaS tool, the solution was to just… use the filesystem sync that's already there.

For developers building with Claude, maintaining OpenClaw skill libraries, or running multi-step AI automations across devices, this is a genuinely practical setup worth adopting. It keeps your agent context consistent, your skills accessible, and your workflow frictionless — without adding any new dependencies to your stack.

The "weird" solutions are often the best ones.

---

Want more Claude automation tips, OpenClaw skill tutorials, and AI DevOps workflows? Explore more at ClawList.io — your hub for practical AI engineering resources.

Original idea credit: @dontbesilent on X