| name | openclaw-message |
| description | Send a message to Michael's OpenClaw / "puffin" agent (the WhatsApp/Telegram bot running on EC2) and get its reply back. Use whenever Michael wants to talk to, ping, test, ask, or relay something to his openclaw/puffin/clawdbot agent, or check whether the agent is responding — e.g. "ask my openclaw bot to…", "ping puffin", "is the agent replying?", "have my bot draft…". Also use to message a specific persona agent (openclaw/michael, openclaw/art, etc.). |
Messaging the OpenClaw agent
This sends a chat message to Michael's OpenClaw agent on EC2 and prints the reply.
How it works (and why it's indirect)
The bot's public OpenAI-compatible endpoint is currently broken — the puffin
wrapper (apps/wrapper/src/server.js, port 80 → http://34.236.245.154/v1/...)
hangs on every completion. The gateway behind it is healthy, but it only
listens on 127.0.0.1:18789 inside the container and its auth token is
regenerated on every restart. So the script reaches the gateway directly via AWS
SSM: it reads the current token from the gateway's startup log, then pipes the
request through docker exec to 127.0.0.1:18789.
If the wrapper proxy ever gets fixed (or the image rebuilt), prefer the public
/v1/chat/completions endpoint instead — it's simpler. Until then, use this.
Usage
python3 scripts/send.py "your message here"
python3 scripts/send.py "draft a reply declining politely" --agent openclaw/michael
python3 scripts/send.py "ping" --raw
- Default agent/model is
openclaw/default. Other personas: openclaw/michael,
openclaw/michael-bot, openclaw/art, openclaw/miranda, openclaw/founders.
- The script handles the SSM round-trip and prints just the assistant's text
(use
--raw for the full response).
- Responses can take 20–90s (the agent runs a full turn), so be patient.
Prerequisites / facts
- AWS account
636160605813, region us-east-1, instance i-02cef0acfa720d57b
(Name puffin-openclaw, container puffin-openclaw-1).
- Uses the custom SSM document
PuffinRunShell (this account has AWS-managed
SSM docs stripped, so a custom run-shell document was created). If SSM calls
fail with InvalidDocument, recreate it:
aws ssm create-document --name PuffinRunShell --document-type Command --document-format JSON --content '{"schemaVersion":"2.2","description":"run shell","parameters":{"commands":{"type":"StringList"}},"mainSteps":[{"action":"aws:runShellScript","name":"run","inputs":{"runCommand":"{{ commands }}"}}]}'
- The agent's default model is
openai-codex/gpt-5.4 (codex OAuth) with
amazon-bedrock/...opus... as fallback.
Troubleshooting
- Empty reply / timeout: the agent turn may be slow or wedged. Check the
container:
aws ssm send-command --region us-east-1 --instance-ids i-02cef0acfa720d57b --document-name PuffinRunShell --parameters 'commands=["docker logs --tail 30 puffin-openclaw-1"]'.
401: the gateway token rotated mid-call; just retry (the script re-reads it each run).- The WhatsApp channel is logged out (
openclaw channels login needed) — unrelated to API messaging.
