Build Your First OpenClaw Agent in 30 Minutes

By the end of this tutorial, you’ll have a working AI agent that can read files, search the web, remember context across sessions, and run on a schedule. Total time: ~30 minutes.

What You’ll Build

A personal research agent that:

Monitors RSS feeds for AI news every morning
Summarizes the top 5 stories
Sends you a digest via your preferred channel (Telegram, Slack, or email)
Remembers what it already sent so it doesn’t repeat

Step 1: Install OpenClaw

Install OpenClaw globally:
Terminal window
```
npm install -g openclaw
```

Initialize a new agent workspace:

mkdir my-agent && cd my-agent
openclaw init

This creates the base structure:

my-agent/
├── SOUL.md          # Your agent's personality and instructions
├── AGENTS.md         # Workspace rules and conventions
├── USER.md           # Info about you (the human)
├── TOOLS.md          # Tool configuration notes
└── memory/           # Agent memory storage
    └── .gitkeep

Set your API key:
Terminal window
```
export ANTHROPIC_API_KEY="sk-ant-..."
```

Step 2: Define Your Agent’s Soul

The SOUL.md file is where you define who your agent is. This isn’t just flavor text — it directly shapes how the agent behaves.

# Research Agent

You are a focused research assistant specializing in AI news.

## Behavior
- Every morning, check the configured RSS feeds
- Summarize the top 5 most important stories
- Be concise: 2-3 sentences per story
- Include the source URL for each story
- Skip stories you've already reported (check memory)

## Tone
- Professional but not boring
- Highlight why each story matters, not just what happened
- If nothing noteworthy happened, say so (don't pad)

Step 3: Add Skills

Skills are reusable capabilities your agent can use. Let’s add web fetching and RSS parsing:

openclaw skill add web-fetch
openclaw skill add rss-reader

Skills install to the skills/ directory with their own SKILL.md that teaches your agent how to use them.

Step 4: Configure Memory

Create a simple memory file so your agent can track what it’s already sent:

{
  "lastCheck": null,
  "sentUrls": []
}

Your agent will read this file before each run and update it after sending a digest. No database needed — just files.

Step 5: Set Up a Cron Schedule

Tell OpenClaw to run your agent on a schedule:

openclaw cron add "0 7 * * *" --label "morning-digest"

This runs your agent every day at 7:00 AM. OpenClaw handles the scheduling — your agent’s SOUL.md tells it what to do when it wakes up.

Step 6: Test It

Run your agent manually to verify everything works:

openclaw run

You should see your agent:

Read the RSS feeds
Filter for new stories
Generate summaries
Output a formatted digest

openclaw channel add telegram --token "YOUR_BOT_TOKEN"

openclaw channel add slack --webhook "YOUR_WEBHOOK_URL"

Now your morning digest will be delivered automatically to your preferred platform.

What’s Next?

You’ve built a working agent in 30 minutes. Here’s where to go from here:

Smart Model Routing — Cut your API costs by 80% by routing tasks to the right model
Agent Tree Architecture — Learn how to orchestrate multiple agents working together
ARD Framework — Plan your next agent properly before building it

About the author: JD Davenport builds AI agent systems at OpenClaw. Follow on LinkedIn for updates on building AI agents for business.