Chapter 5: Communication Automation
The message tool is OpenClaw's interface to messaging channels—WhatsApp is the primary integration. Basic sends require a target (phone number with country code or saved contact name) and a message string, with optional media for attachments like PDFs or images. WhatsApp formatting supports bold, _italic_, ~strikethrough~, and ` monospace , with \n` for multi-line messages. Outbound automation is the simpler pattern: set up cron jobs that compile briefings and message at the right times—morning briefings with weather, task lists, and log alerts; periodic health checks that message only when something is wrong (proactive silence is a design goal).
Inbound WhatsApp automation turns OpenClaw into a conversational AI contact in your phone. When you message it, it creates an agent session with full tool access, reading routing rules from SOUL.md. Keyword routing patterns in SOUL.md direct messages: "status" triggers log checks, "report" generates project reports, "check [url]" fetches and summarizes URLs, "remind me [time] [thing]" sets reminders. The tts (text-to-speech) tool delivers voice briefings and hands-free alerts, useful when you can't look at your phone. After TTS, the agent should reply with NO_REPLY to avoid duplicate text delivery.
A production alert system has three severity levels: CRITICAL (immediate WhatsApp + TTS, e.g., pipeline failure, disk >90%), WARNING (WhatsApp during daytime, logged at night, e.g., disk 80-90%, service degraded), and INFO (log file only, e.g., normal completions). Quiet hours (10pm–7am) suppress non-critical alerts. The alert dispatcher is a Python script that categorizes by level, respects quiet hours, and routes to appropriate channels. Privacy considerations: never send credentials or PII via WhatsApp, be aware of third-party message retention, and ensure compliance (CAN-SPAM, GDPR) when automating messages to others. The communication hub architecture combines data sources, processors, routing rules, and channels into an intelligent system where proactive silence (no message when everything is normal) is as important as proactive alerts.
Key Patterns
- **Message formatting:** `*bold*`, `_italic_`, `~strikethrough~`, `` `monospace` `` with `\n` for newlines
- **Keyword routing in SOUL.md:** Messages starting with "status", "report", "check [url]", "remind me" route to specific workflows
- **TTS with NO_REPLY:** Use text-to-speech for voice briefings; reply with NO_REPLY to suppress duplicate text
- **Three-tier alert system:** CRITICAL (WhatsApp+TTS immediately), WARNING (WhatsApp daytime only), INFO (log only)
- **Proactive silence:** Don't send "system normal" notifications; only alert when something actually needs attention
Related Concepts
- [[kelly-handbook-ch6-cron]] for scheduling message delivery and periodic health checks
- [[kelly-handbook-ch2-architecture]] for understanding channels as part of the Gateway pipeline
- [[kelly-handbook-ch3-file-automation]] for the file pipelines that feed data into alerts
- [[kelly-tweets-openclaw]] for OpenClaw messaging and channel patterns