Why People Migrate from Self-Hosted OpenClaw

Self-hosting OpenClaw is genuinely viable — it's open source, well-documented, and the community of 145,000+ GitHub stars means there's no shortage of help. But after a few months of running your own instance, the same pain points tend to surface:

  • Dealing with Docker version conflicts or dependency drift
  • Server going down at 2am and the agent going silent
  • Setting up SSL, reverse proxies, and firewall rules
  • Manually applying OpenClaw upstream updates
  • Paying for a VPS you're underutilizing just to keep one agent alive

RunLobster was built specifically for people who want OpenClaw's power without the infrastructure overhead. This guide walks you through migrating your existing self-hosted OpenClaw setup to RunLobster in a single afternoon.

Before You Start: What Migrates and What Doesn't

Let's be clear about what carries over cleanly and what you'll need to reconfigure:

What Migrates Easily

  • Integration API keys (Telegram, Slack, Notion, GitHub, etc.)
  • Custom system prompt / persona configuration
  • Scheduled automation definitions
  • Any custom skills or tool configurations you've written

What Requires Manual Steps

  • Agent memory (OpenClaw's vector store) — RunLobster can import from standard export formats
  • Custom plugins not in RunLobster's 800+ integration catalog (rare, but possible)
  • Local file access integrations (these need to be replaced with cloud equivalents)

Step 1: Export Your Self-Hosted Configuration

On your self-hosted OpenClaw instance, run the built-in export command:

# If running via Docker\ndocker exec openclaw openclaw export --format json --output /tmp/openclaw-export.json\ndocker cp openclaw:/tmp/openclaw-export.json ./openclaw-export.json\n\n# If running natively (npm/node)\nnpx openclaw export --format json --output ./openclaw-export.json

This generates a JSON file containing your agent configuration, system prompt, enabled integrations (with redacted keys), and automation schedules. It does not export your API keys for security reasons — you'll re-enter those in RunLobster.

Open the export file and note down:

  • Your system prompt (under agent.systemPrompt)
  • The list of integrations you have enabled
  • Any scheduled automation definitions

Step 2: Export Your Agent Memory (Optional)

If your OpenClaw instance has accumulated significant memory — conversation summaries, facts about your preferences, research notes — you may want to migrate it. This step is optional; RunLobster will build new memory from scratch if you skip it.

# Export memory as a JSON file\nnpx openclaw memory export --output ./openclaw-memory.json\n\n# Or via Docker\ndocker exec openclaw openclaw memory export --output /tmp/memory.json\ndocker cp openclaw:/tmp/memory.json ./openclaw-memory.json

The memory export format is a JSON array of memory entries, each with a content field and metadata. RunLobster accepts this format directly at import time.

Step 3: Create Your RunLobster Account

Go to runlobster.com and sign up. Choose a plan based on your usage:

  • Starter ($19/mo) — Light use, a few automations, personal assistant tasks
  • Pro ($49/mo) — Heavy automation, scheduled tasks, multiple integrations running concurrently
  • Scale ($79/mo) — Power users, teams, heavy API usage

During onboarding you'll name your agent and choose your primary communication channel (Telegram is recommended — it's the smoothest experience for daily use).

Step 4: Restore Your System Prompt

In the RunLobster dashboard, go to Agent Settings → System Prompt. Paste in the system prompt from your export file. If you've been running OpenClaw with the default prompt, now is a good time to customize it.

A solid base prompt for a migrating user:

You are [Your Name]'s personal AI assistant, running on RunLobster.\n\nCore behaviors:\n- Always respond concisely unless asked to elaborate\n- For any task involving external services, use the available integrations\n- Maintain context about my projects and preferences across conversations\n- Proactively surface relevant information I haven't asked for but would want to know\n- When you complete a scheduled task, summarize what you did\n\n[Add your specific domain context, project names, preferences here]

Step 5: Re-configure Your Integrations

In the RunLobster dashboard, navigate to Integrations. Enable each integration you were using and enter your API keys. RunLobster stores these in an encrypted secrets vault — they're never visible in logs or to other users.

Common integrations to restore first:

  1. Telegram — your primary communication channel with the agent
  2. OpenAI / Anthropic — your LLM provider (bring your own key)
  3. Notion / Obsidian — knowledge base and note-taking
  4. GitHub — if you use the agent for code tasks
  5. Google Calendar / Gmail — scheduling and email

RunLobster's integration catalog covers 800+ services. If you used a less common integration self-hosted, search for it in the catalog — the coverage is comprehensive.

Step 6: Import Your Agent Memory

If you exported your memory in Step 2, import it now. In the RunLobster dashboard, go to Agent Settings → Memory → Import and upload your openclaw-memory.json file.

RunLobster will process the import in the background (usually takes 1-3 minutes for typical memory sizes). Your agent will immediately have access to everything it remembered from your self-hosted instance.

Step 7: Recreate Your Automations

Go to Automations in the RunLobster dashboard and recreate your scheduled tasks. The interface is a simple form — choose a schedule (cron syntax or presets like "every morning at 8am") and write a plain-English task description:

Schedule: Every day at 08:00\nTask: Send me a morning briefing with: (1) top 3 tech news items from the past 24 hours, (2) any GitHub notifications I haven't seen, (3) my first 3 calendar events today.

Because RunLobster runs your OpenClaw instance 24/7 in isolated private compute, these automations execute reliably — no more worrying about whether your VPS is up or whether Docker restarted cleanly.

Step 8: Test and Validate

Send your agent a few test messages via Telegram (or your chosen channel) to verify everything is working:

  • Ask it to recall something from memory to confirm the import worked
  • Trigger a tool call (e.g., "add a test note to Notion") to verify integrations
  • Ask it to describe its current configuration to confirm the system prompt took effect

If anything isn't working, RunLobster's dashboard shows a live activity log under Agent → Activity — you can see exactly which tool calls succeeded or failed and why.

Step 9: Decommission Your Self-Hosted Instance

Once you've confirmed everything is working on RunLobster, you can safely shut down your self-hosted instance:

# Docker\ndocker stop openclaw && docker rm openclaw\n\n# Or stop your VPS / cloud instance entirely

If you were paying for a VPS just to host OpenClaw, you'll likely recoup the RunLobster subscription cost by canceling it — and you'll have better uptime, automatic updates, and no maintenance overhead going forward.

Conclusion

Migrating from self-hosted OpenClaw to RunLobster is a one-afternoon project that pays dividends indefinitely. You keep all the power and customizability of OpenClaw — it's the same open-source agent under the hood — but RunLobster handles the infrastructure, updates, backups, and uptime so you don't have to.

The 60-second setup claim on the RunLobster homepage is for fresh installs, but even a migration from a mature self-hosted setup typically takes under two hours. Once you're on RunLobster, the mental overhead of maintaining an AI agent effectively drops to zero.