Migrate from OpenClaw
If you're coming from an OpenClaw setup, agentx migrate imports the bulk of your config in one shot.
Command
bash
agentx migrate openclaw # reads default OpenClaw config path
agentx migrate openclaw ./openclaw.json # explicit path
agentx migrate openclaw --dry-run # preview without writingThe migrator produces (or updates) agentx.json in the current directory and scaffolds Claude Code workspaces for each imported agent.
What gets imported
| OpenClaw | AgentX equivalent |
|---|---|
| Agents + personalities | agents.<id> + bootstrap files in the agent's workspace |
| Telegram accounts | channels.telegram.accounts |
| Channel-agent bindings | channels.*.agentBinding / mentions |
| Scheduled jobs | crons.<id> with cron syntax |
| Project metadata | node.name, agent workspace descriptions |
What doesn't auto-migrate
- Secrets — tokens are referenced as
${TG_…_BOT_TOKEN}; you populate.envmanually - Custom OpenClaw plugins — rewrite as AgentX hooks or MCP servers
- Business-layer style rules — re-author under
business.*inagentx.json(see Journey 7)
After migration
- Populate
.envwith the tokens the migrator listed - Run
agentx config check— it surfaces any missing references - Start the daemon:
agentx daemon start - Watch:
agentx daemon watch
Rollback
The migrator writes a timestamped backup agentx.json.bak.<timestamp> before overwriting. Restore with:
bash
cp agentx.json.bak.<timestamp> agentx.json