How to use the Migrate from OpenClaw agent skill

What the skill makes the agent do

Once installed, the skill lets an AI agent act as a guide for a migration. It first scans your machine for an OpenClaw install, then summarizes what it found, such as the assistant name, channels, scheduled jobs, workspace files and MCP servers. It then talks through how each piece should map onto NanoClaw v2's structure.

The skill is built as a conversation rather than a bulk copy. It reads OpenClaw state, explains it, places it, and shows proposed changes before applying them. Credentials are masked when displayed, showing the first 4 characters, three dots, and the last 4.

• Detects an OpenClaw installation at standard locations or a custom path you give it
• Extracts identity, channel credentials, scheduled tasks and other config
• Asks multiple-choice questions for decisions and free-form questions for names and details
• Applies changes only after you agree to them

What you need before you start

The skill has several requirements. Check them before you install, because the migration cannot run without them. The skill also runs scripts on your machine, so this is worth understanding first.

• Node.js with pnpm, since the scripts run through pnpm exec tsx
• An existing NanoClaw v2 checkout to migrate into
• An OpenClaw installation or its state directory, by default ~/.openclaw or ~/.clawdbot
• The OneCLI CLI, used to store container-facing API credentials in the Agent Vault
• The NanoClaw service running when the owner and primary agent are seeded, because the setup queues a welcome message through the running service

How to install the skill

Download the ZIP of the skill from this page and unzip it. Then place the skill folder in one of two locations. Use ~/.claude/skills/ to make it available for your user account, or use a project's .claude/skills/ folder to make it available only inside that project.

If you use the Claude apps instead, you can upload the ZIP there. Open Settings, go to Capabilities, then Skills, and upload the skill from that screen.

• Option A: unzip into ~/.claude/skills/ for personal use
• Option B: unzip into your project's .claude/skills/ folder for one project
• Option C: upload the ZIP in the Claude apps under Settings, Capabilities, Skills

How to trigger the skill with natural requests

You do not need a special command. The skill responds to plain requests that contain its trigger phrases. Use one of the phrases below in a conversation with your agent, and the skill should start the discovery step.

If you have OpenClaw in a non-standard location, say so in the same message. The discovery script accepts a custom state directory, and the agent will pass it along when you give it a path.

• "migrate from openclaw"
• "openclaw migration"
• "import from openclaw"

What happens during a migration

The migration runs in numbered phases, and the agent keeps a progress file in your project root called migration-state.md. The file records what was found, what you decided, and which phases are done. If the conversation is interrupted, the agent can re-read this file to recover its place.

The first phase is discovery. If no OpenClaw install is found, the agent asks for a custom path and stops if you do not give one. If an install is found, the agent summarizes it and asks whether you are ready to continue. Later phases cover agents and groups, the assistant name, and the owner and primary agent.

• Phase 0 runs discovery and summarizes what was found
• Phase 1 decides whether groups share one identity or are fully separate, then confirms the assistant name
• Later phases move credentials, memory, identity files and scheduled tasks, with a review before each change is applied
• Phase 8 validates the install and offers to delete migration-state.md or keep it as a record

Safety notes: scripts, network and secrets

This skill runs TypeScript scripts on your machine and reads config files that contain credentials. The source states that it does not use the network. You should still read the scripts and the migration plan before approving changes, especially for channel tokens.

Container-facing API keys such as Anthropic or OpenAI are stored in the OneCLI Agent Vault. Host-side channel tokens for Telegram, Discord or Slack stay in .env, because the NanoClaw host process needs them to connect. Masked display is for showing values in the conversation; it does not change where secrets are stored.

• Scripts run: yes. Uses network: no
• Credentials are shown masked, first 4 characters, three dots, last 4
• Do not use this skill if you do not want scripts run on your machine or secrets handled by an agent
• Review each proposed change before agreeing to it

Tips for a smooth migration

Start with a copy of your OpenClaw state directory if you want a fallback, and make sure your NanoClaw v2 checkout is in a clean state before you begin. Answer the shared-versus-separate question deliberately, because it decides where files go. If the discovery output seems to miss something, ask the agent to read that part of the config directly.

To remove what the skill copied into your project, follow REMOVE.md. Note that REMOVE.md deletes the files the skill installs, and it does not undo the agents, groups, roles, .env tokens or vault secrets created by the migration. Check the file names it lists against the skill's file list before running any removal command.

• Keep an OpenClaw backup before running the migration
• Decide shared versus separate identity before any file is written
• Ask the agent to read sections of config directly if discovery reports unfamiliar keys
• Scheduled tasks become task rows in v2 with cron recurrence, so review the mapped schedules