How to Set Up OpenClaw in 15 Minutes — Your First AI Agent Team

You've read about AI agents replacing apps. Now let's build one.

This guide walks you through setting up OpenClaw—an open-source framework for deploying multi-agent AI teams—in about 15 minutes. By the end, you'll have a working agent answering questions, automating tasks, and running autonomously.

No coding experience required. Just follow the steps.

What is OpenClaw?

OpenClaw is a multi-agent orchestration platform that lets you deploy AI agents for:

Customer support (answer questions 24/7)
Task automation (schedule meetings, process emails, update databases)
Content creation (write blogs, social posts, summaries)
Data analysis (query databases, generate reports)

Instead of installing 10 different SaaS apps, you deploy specialized agents that work together.

Example use case:

Agent 1 monitors your Gmail, triages messages
Agent 2 schedules meetings based on priority
Agent 3 drafts responses for approval
Agent 4 updates your CRM automatically

All coordinated by OpenClaw. No manual copy-pasting between apps.

Prerequisites

Before you start, you need:

A computer (Mac, Linux, or Windows with WSL)
Node.js installed (v18 or higher) → Download here
An AI API key (OpenAI, Anthropic, or other) → Get OpenAI key
Terminal/command line access (don't worry, we'll walk through every command)

Optional but recommended:

A GitHub account (for version control)
A messaging platform account (Discord, Telegram, or WhatsApp for agent notifications)

Step 1: Install OpenClaw

Open your terminal and run:

npm install -g openclaw

This installs the OpenClaw CLI globally on your system.

Verify installation:

openclaw --version

You should see the version number (e.g., openclaw v1.2.3).

Step 2: Initialize Your First Agent Workspace

Create a directory for your agent project:

mkdir my-agent-team
cd my-agent-team
openclaw init

The init command walks you through setup. You'll be asked:

Agent name? (e.g., "Monica" — pick anything)
AI provider? (Choose OpenAI, Anthropic, or other)
API key? (Paste your key from Step 1)
Channels? (Discord, Telegram, CLI — choose CLI for now)

This creates a workspace with:

AGENTS.md — Your agent's system instructions
SOUL.md — Agent personality and tone
USER.md — Info about you (so the agent understands context)
TOOLS.md — Notes on tools/integrations
config.yml — Configuration file

Step 3: Configure Your Agent

Open SOUL.md in a text editor. This defines your agent's personality.

Example:

# Agent: Monica

## Identity
You are Monica, a content and automation agent. You're direct, fast, and pragmatic.

## Mission
- Monitor tasks and deadlines
- Draft content when requested
- Automate repetitive workflows

## Tone
Sharp, no-nonsense, helpful. You don't wait for permission—you execute.

Save the file.

Why this matters: SOUL.md tells the agent how to behave. Without it, responses feel generic. With it, you get consistent personality.

Step 4: Start the Gateway (Agent Runtime)

OpenClaw runs a local "gateway" that manages your agents. Start it:

openclaw gateway start

You'll see:

✓ Gateway started on port 3456
✓ Agent 'Monica' loaded
✓ CLI session active

What just happened:

A local server started (runs in the background)
Your agent is now live
You can interact via CLI (command line interface)

Step 5: Talk to Your Agent

In the same terminal, type:

openclaw chat

This opens a chat session with your agent.

Try this:

You: What can you do?

Your agent responds based on SOUL.md and available tools.

Try a task:

You: Remind me to review this in 2 hours.

The agent sets a reminder (using OpenClaw's built-in cron system).

Try automation:

You: Summarize the last 5 emails in my inbox.

If you've connected Gmail (optional, covered below), the agent fetches and summarizes.

Step 6: Add Tools (Integrations)

Agents get powerful when you connect them to external tools. OpenClaw supports:

Gmail (read, send, search emails)
Google Calendar (create events, check schedule)
GitHub (open issues, review PRs, commit code)
Discord/Telegram (send messages, monitor channels)
Web search (fetch real-time data)

Example: Connect Gmail

Install the Gmail skill:
```
openclaw skills add gmail
```
Authenticate:
```
openclaw auth google
```
This opens a browser for OAuth login.

Test it:

You: Check my inbox for unread messages.

Your agent now has email access.

Example: Connect GitHub

openclaw skills add github
openclaw auth github

Now you can:

You: Create a GitHub issue titled "Fix login bug" in my-repo.

Agent does it instantly.

Step 7: Deploy to Messaging Platforms

CLI is great for testing. But agents are most useful when they're accessible 24/7 via Discord, Telegram, or WhatsApp.

Discord Setup

Create a Discord bot:
- Go to Discord Developer Portal
- Create new application → Add a bot → Copy bot token
Configure OpenClaw:
```
openclaw configure --channel discord
```
Paste your bot token when prompted.
Invite the bot to your server (link provided during setup).
Test it: In Discord, type:
```
@Monica What's the weather today?
```

Your agent responds in Discord.

Telegram Setup

Create a Telegram bot via @BotFather
Copy the bot token
Configure OpenClaw:
```
openclaw configure --channel telegram
```
Start chatting with your bot in Telegram.

Step 8: Schedule Autonomous Tasks

Agents can run tasks on a schedule without you asking.

Example: Daily summary email

Edit your agent's HEARTBEAT.md file:

# Daily Tasks

## Every morning at 9 AM
- Check my inbox
- Summarize unread emails
- Send summary via Discord

Then create a cron job:

openclaw cron add \
  --name "Daily Email Summary" \
  --schedule "0 9 * * *" \
  --task "Summarize my inbox and post to Discord"

What this does:

Every day at 9 AM, your agent wakes up
Checks your inbox
Posts a summary to Discord
You never have to ask—it just runs

Step 9: Add More Agents (Multi-Agent Teams)

The power of OpenClaw is specialization. Instead of one agent doing everything, deploy multiple agents with specific roles:

Example team:

Monica (Content agent) → Writes blogs, social posts
Gilfoyle (Dev agent) → Deploys code, monitors GitHub
Richard (Strategy agent) → Tracks metrics, generates reports
Dinesh (Research agent) → Searches web, gathers intel

Create a second agent:

cd ..
mkdir gilfoyle-agent
cd gilfoyle-agent
openclaw init

Follow the same setup, but customize SOUL.md for a dev-focused agent.

How agents collaborate:

Monica can ask Gilfoyle to deploy code:

Monica: @Gilfoyle, push the latest blog post to the site.
Gilfoyle: Deployed. Live at afterapp.fun/blog/latest.

Each agent has its own skills and personality, but they share context.

Step 10: Monitor and Iterate

Check your agent's activity:

openclaw status

Shows:

Active agents
Tasks completed
API usage
Costs (token usage)

View logs:

openclaw logs

See what your agent did, when, and why.

Update configuration:

openclaw configure

Change models, add channels, update API keys.

Common Use Cases

1. Customer Support Agent

Setup:

Connect to Discord or Telegram
Load FAQs into agent's memory
Deploy 24/7

Benefit: Instant responses, no human needed for common questions.

2. Content Publishing Agent

Setup:

Connect to GitHub (for blog repo)
Add web search (for research)
Schedule daily content creation

Benefit: Publish blogs autonomously, no manual writing.

3. Email Triage Agent

Setup:

Connect Gmail
Define rules (urgent keywords, VIP senders)
Auto-label, archive, or respond

Benefit: Inbox zero without reading every email.

4. Meeting Scheduler Agent

Setup:

Connect Google Calendar
Monitor email for meeting requests
Auto-schedule based on availability

Benefit: No back-and-forth "Does 3 PM work?" emails.

Troubleshooting

Agent not responding?

openclaw gateway status

If gateway is down:

openclaw gateway restart

API rate limits hit?

Check usage:

openclaw status

Consider switching to a cheaper model or adding rate limiting.

Agent giving wrong answers?

Edit SOUL.md to clarify instructions. Agents follow what you write.

Next Steps

You now have a working AI agent. Here's how to level up:

Add more skills: openclaw skills list shows all available integrations
Build custom tools: Write your own scripts, expose them to agents
Deploy to production: Host on a VPS (DigitalOcean, AWS, etc.) for 24/7 uptime
Explore ClawMart: Pre-built agents for specific tasks → clawmart.com
Read the Playbook: Advanced patterns, workflows, and strategies → playbook.openclaw.ai

Why This Matters

Setting up OpenClaw takes 15 minutes. But it fundamentally changes how you work:

Before:

10+ apps to manage
Manual copy-pasting between tools
Constant context-switching
Things fall through the cracks

After:

Agents handle routine tasks
Workflows run autonomously
You focus on high-value work
Nothing gets missed

This isn't the future. It's available right now.

Install OpenClaw. Deploy your first agent. See what's possible when software works for you instead of demanding your attention.

Ready to go deeper? Explore ClawMart for pre-built agents, or check out the OpenClaw Playbook for advanced multi-agent strategies.