An AI agent that messages you when things need your attention — built in an afternoon.

Overview

Install OpenClaw, connect a messaging channel, configure a useful read-only skill, and evaluate whether it’s worth keeping.

Important: Before starting, read Module 05’s updated security section. The threat landscape for OpenClaw has escalated significantly since this course launched. Docker sandboxing and the security configuration steps below aren’t optional — they’re essential.

Draws from: Module 04 Time estimate: 2-3 hours Difficulty: Medium Prerequisites: Node.js 22+, Docker, a Telegram account (easiest first channel)

Step-by-Step

Phase 1: Install and Configure (45-60 min)

1. Install OpenClaw:

   npm install -g openclaw
   

2. Run the onboarding wizard:

   openclaw onboard
   

3. During onboarding:
   • Select your AI provider (Anthropic recommended if you have a key)
   • Create a Telegram bot via @BotFather (follow the wizard prompts)
   • Enable the core/daily-digest skill
4. Configure security (per Module 05):
   • Set confirmBeforeAction: true
   • Set sandboxMode: "docker"
   • Enable audit logging

Phase 2: Configure Your First Skill (30-45 min)

Start with a daily digest — low risk, immediate value:

1. Configure data sources (email via IMAP, or calendar via API)
2. Set the schedule (e.g., 8:00 AM daily)
3. Set the output channel (Telegram)
4. Set the format (brief summaries, urgency-flagged)

Alternatively, start with a document summarizer:

1. Enable the core/summarize skill
2. Send a PDF or long text to your Telegram bot
3. Get a summary back

Phase 3: Test and Evaluate (45-60 min)

1. Send test messages through your connected channel
2. Check responses for accuracy:
   • Does the summary capture key points?
   • Does it miss anything important?
   • Does urgency flagging make sense?
3. Review the audit log: what actions did the agent take? Any surprises?
4. The evaluation question: Did this save me time today? Would it save time every day?

What Success Looks Like

• A running OpenClaw instance with one functioning skill
• A clear sense of whether autonomous agents are useful for your workflow right now
• If not: you’ve learned the technology and can revisit when your use case clarifies
• If yes: a foundation to add more skills gradually

Stretch Goals

• Add a second messaging channel (Discord, WhatsApp)
• Configure a second skill (document summarization, web search)
• Set up the trust escalation: move from read-only to draft mode on one skill

Previous: Project B: Build a Personal Data Dashboard | Back to Project Lab | Next: Project D: Build an n8n Workflow