How to Migrate Your OpenClaw from Self-Hosted to Managed

Made the decision to move from a self-hosted VPS to nacre.sh? Smart choice. This guide walks through the complete migration process — exporting your configuration, skills, and memory from your existing instance, and importing it cleanly into nacre.sh. Total migration time is typically 20–40 minutes.

What Can Be Migrated

✅ Migrates cleanly:

• OpenClaw configuration (openclaw.json)
• Installed skills list and their configurations
• Agent memory and conversation history
• Custom system prompts and persona settings
• Channel configurations (Telegram bot token, Discord token, etc.)

⚠️ Requires manual attention:

• API keys (you'll re-enter these in nacre.sh's secure key store)
• Custom skill files you've written locally
• Secrets stored in environment variables

❌ Cannot migrate:

• Local files your agent had access to (filesystem access scope differs between self-hosted and managed)
• Database connections to localhost services

Step 1: Export from Self-Hosted

SSH into your self-hosted instance and create an export:

cd ~/openclaw  # or wherever OpenClaw is installed

# Export full configuration
docker compose exec openclaw openclaw export --output /tmp/openclaw-export.zip

# Copy the export file to your local machine
# On your LOCAL machine:
scp user@your-server-ip:/tmp/openclaw-export.zip ./openclaw-export.zip

The export file includes your configuration, memory, and skills list (but not API keys for security reasons).

Step 2: Record Your Current Settings

Before migrating, document:

# List installed skills
docker compose exec openclaw openclaw skills list

# View current configuration (will show non-sensitive settings)
docker compose exec openclaw openclaw config show

Screenshot or note:

• All installed skills and their versions
• Telegram/Discord bot tokens (you'll need these in nacre.sh)
• Any custom system prompt text
• LLM model selection and any custom parameters

Step 3: Create Your nacre.sh Instance

1. Go to nacre.sh/signup and choose a plan
2. Select your preferred region (closest to your users)
3. Your instance will be provisioned in approximately 2–3 minutes

Step 4: Import Your Configuration

In your nacre.sh dashboard:

1. Click Import Configuration
2. Upload your openclaw-export.zip file
3. The platform imports your skills list, memory, and configuration automatically

Step 5: Re-enter API Keys and Tokens

In your nacre.sh instance settings, re-enter:

• LLM API key (Anthropic, OpenAI, etc.)
• Telegram bot token
• Any other channel bot tokens
• Skills that require their own API keys (Google OAuth, GitHub token, etc.)

nacre.sh stores all keys in encrypted vaults, isolated per customer.

Step 6: Verify and Test

Before shutting down your old server:

1. Test your agent responds correctly in your connected channel
2. Verify memory recall works ("What did we talk about yesterday?")
3. Test each installed skill with a simple command
4. Confirm your custom system prompt is active

Step 7: Decommission the Old Server

Once you're satisfied the migration is complete:

# On your old server — graceful shutdown
docker compose down

# Cancel your VPS subscription
# The server will be deleted at the end of your billing period

Keep your openclaw-export.zip for at least 30 days as a backup.

Frequently Asked Questions

Will my Telegram/Discord bot users notice any disruption?

If you migrate during low-usage hours, the transition is seamless. The bot token stays the same — users continue chatting without any visible change.

Can I run both old and new instances simultaneously?

For Telegram and WhatsApp, only one instance can use a given bot token at a time. Briefly disconnect the channel on your old instance before connecting it to nacre.sh.