Migrating to Your Agent

If you've been running an AI assistant somewhere else — on a laptop, a server, another platform — you don't have to start from zero. Your AI Agent ships with a built-in migration skill that moves your old assistant's accumulated life onto the appliance: the conversations, the memory, the files, the skills, the schedules. What took months to build up arrives in one guided session.

What migration brings across

That split is the point: migration brings your data and know-how, not your old plumbing. Everything the turnkey appliance already handles — models, connectivity, backups, the web interface — stays the appliance's pre-configured setup, so you land with your history and none of your old maintenance burden.

How it works

The whole flow runs as a conversation with your new agent:

1. Ask the agent to migrate. Something like "I want to migrate my old assistant from my Mac" starts the skill. It asks a couple of questions about where the old assistant lives.
2. Run one line on the old computer. The agent gives you a single command to run on the machine that holds your old assistant. You run it in a terminal — the Terminal app on macOS and Ubuntu, or PowerShell on Windows — then simply leave that window open until the migration finishes. That's the entire footprint on the old side: nothing to install. Never used a terminal before? See Opening a terminal for step-by-step instructions and screenshots for Windows, macOS, and Ubuntu.
3. The agent scans and proposes a plan. It examines what it found — how many conversations, how much memory, which files, which skills — and presents a migration plan for your approval. Nothing moves until you approve it.
4. Transfer runs, with resume. Data moves across encrypted; if the connection drops mid-way, the transfer resumes where it left off rather than starting over.
5. Install and hand-back. The agent installs what arrived — history, memory, files, skills, schedules — then walks you through a short setup pass to confirm everything landed and reconcile anything (like the personality) you may want to blend with what you'd already set up.

How the data travels

The transfer is peer-to-peer: your old computer and your AI Agent Processor talk to each other over a secure tunnel handled by Cloudflare — one of the world's largest internet infrastructure companies, whose global network of data centres sits in front of a significant share of the web. That global footprint is why the transfer is so dependable:

• Same network? If the old computer and the device are on the same network, the transfer is usually direct between the two machines — the data often never leaves your building.
• Different networks? The two sides negotiate a direct peer-to-peer connection across the internet where possible.
• Direct connection blocked? Where a direct connection isn't possible (strict firewalls, carrier-grade NAT), the data automatically relays through the closest possible Cloudflare node instead of failing.

You don't choose between these — the connection always finds the best available path on its own.

What you can migrate from

The migration skill natively understands a range of popular AI assistants and tools, including:

• Hermes and OpenClaw / Clawbot agents — including a previous agent of the same lineage
• Claude Code and Claude Desktop
• ChatGPT Desktop and Codex (the ChatGPT CLI) — plus official ChatGPT and Claude conversation export archives
• Cursor
• Gemini
• Aider
• GitHub Copilot

The scan step is honest about what it can and can't extract from each source.

And because the whole migration flow is AI native — your agent itself drives the discovery, the plan, and the install — it isn't limited to that list. If your old assistant or data source isn't mentioned above, let the agent do its own discovery: tell it where the old assistant lives, then prompt it and work with it to make sure it finds and brings across everything you need.

Migrating just part of it

You don't have to take everything. The migration plan is itemised and you choose what comes across — just the conversation history, just your skills and scheduled tasks, just the memory and personality, or any combination. That's handy when you want your old assistant's know-how without its clutter, or when you're pulling selected pieces from more than one source.

Before you start

• Take a backup first. If your new agent already has any state you care about, Back up now with a note — migration adds substantial content, and a restore point means you can rewind the whole thing. See Backups.
• Both machines online. The old computer needs to reach the internet for the transfer; the agent handles the rest.
• Budget headroom. Installing a large memory archive involves some model work — worth knowing if you run tight spend caps.

Migration, restore, or replace?

Three tools move agent state around — use the right one:

After the migration

Spend the first day or two the way you would with the onboarding conversation: ask the agent what it remembers, spot-check the migrated memory, skim the workspace file list, and confirm any migrated scheduled tasks still make sense in their new home (they now run in your facility's timezone, on always-on hardware). Then let the hourly backups take over protecting all of it.

One extra step worth doing: ask the agent to re-index everything it migrated into its long-term memory. The agent's memory normally indexes conversations as it lives through them, so an imported archive isn't automatically part of its recall. Ask something like "re-index the migrated history into your long-term memory" — it involves some model work (the agent will confirm before spending the credit), and once done, future conversations are context-aware of everything your old assistant knew.

Related pages

• Opening a terminal — running the migration command on Windows, macOS, and Ubuntu
• Getting started — if this is also a brand-new device
• Skills — the migration skill's siblings
• Backups — the pre-migration safety net