Home/Blog/OpenClaw: The Complete Step-by-Step Guide to Your ...
AI⭐ Featured

OpenClaw: The Complete Step-by-Step Guide to Your Personal AI Agent

Learn how to set up and use OpenClaw — the open-source personal AI assistant that runs on your machine, connects to any chat app, and autonomously gets things done 24/7. A complete guide with real-world examples.

Sani Mridha

Sani Mridha

Senior Mobile Developer

📅 2026-03-20⏱️ 18 min read
🤖

OpenClaw: The Complete Step-by-Step Guide to Your Personal AI Agent

Imagine having a senior developer, personal assistant, and research analyst all rolled into one — available 24/7, running on your own machine, and reachable via WhatsApp. That's OpenClaw.

OpenClaw is an open-source personal AI assistant that actually *does* things. Unlike standard chatbots, it has eyes and hands: it can browse the web, read and write files, run shell commands, and proactively work in the background while you sleep.

Let's walk through everything you need to know to set it up and get the most out of it.

What Makes OpenClaw Different?

It's Not Just a Chatbot

Most AI assistants answer questions. OpenClaw takes action:

  • Browser Control: Fills out health reimbursements, books doctor appointments, scrapes data from any site
  • Terminal Access: Reads/writes files, runs shell commands, executes code in a secure sandbox
  • Proactive Intelligence: Runs background cron jobs and heartbeats, working while you're offline
  • Persistent Memory: Remembers your Obsidian notes, WHOOP metrics, biomarker goals, and personal preferences
  • Local Sovereignty

    Your data stays on your machine — not on a tech giant's server. OpenClaw runs on Mac, Windows, or Linux. Your conversation history, personal context, and sensitive files (Gmail, calendar, local documents) never leave your hardware unless you explicitly configure it that way.

    Multi-Platform Chat Interface

    You don't need a new app. OpenClaw works through the tools you already use:

  • WhatsApp
  • Telegram
  • Discord
  • Slack
  • Signal
  • iMessage
  • Just message it like you'd message a coworker.

    Prerequisites

    Before getting started, make sure you have:

  • A Mac, Windows, or Linux machine
  • Docker installed (for the gateway)
  • Node.js 18+ (for local development)
  • An API key from OpenAI, Anthropic, or another provider (or subscribe to the built-in API plan)
  • Step 1: Install OpenClaw

    OpenClaw provides a one-liner installation script. Open your terminal and run:

    curl -fsSL https://open-claw.org/install.sh | bash

    This script will:

    1. Detect your operating system

    2. Pull the required Docker images

    3. Set up the OpenClaw daemon

    4. Start the gateway service on localhost

    Wait for the script to complete. You should see a success message with a local URL for the dashboard.

    Verify the Installation

    openclaw status

    Expected output:

    ✅ OpenClaw Gateway: Running on localhost:3210
    ✅ Memory Module: Active
    ✅ Browser Engine: Ready
    ✅ Background Worker: Online

    Step 2: Configure Your AI Model

    OpenClaw supports multiple AI providers. You can configure your preferred model in the settings file:

    openclaw config --edit

    This opens your config file at ~/.openclaw/config.yml. Here's a sample configuration:

    model:
      provider: anthropic
      name: claude-opus-4-5
      api_key: YOUR_API_KEY_HERE
    
    gateway:
      port: 3210
      bind: localhost
      ssh_tunnel: true
    
    memory:
      backend: local
      path: ~/.openclaw/memory

    Using the Built-in API (No Key Required)

    If you subscribe to the Pro plan, you get $60 in built-in API credits with access to 5 flagship models:

  • Opus 4.6 — Best for complex reasoning and coding
  • GLM-5 — Excellent for multilingual tasks
  • KIMI K2.5 — Great for document processing
  • GPT-5.2 Codex — Ideal for software development
  • Gemini 3 Pro — Strong at multimodal tasks
  • Using Local Models via Ollama

    For maximum privacy, run fully local models:

    brew install ollama
    ollama pull llama3.1:8b
    
    openclaw config set model.provider ollama
    openclaw config set model.name llama3.1:8b
    openclaw config set model.api_base http://localhost:11434

    Step 3: Connect Your Chat Platform

    Option A: Telegram (Recommended for Beginners)

    1. Open Telegram and search for @BotFather

    2. Send /newbot and follow the prompts

    3. Copy the bot token you receive

    4. Run the connection command:

    openclaw connect telegram --token YOUR_BOT_TOKEN

    5. Start chatting with your new bot!

    Option B: Discord

    1. Go to the Discord Developer Portal and create a new application

    2. Add a Bot and copy the bot token

    3. Add the bot to your server with the right permissions:

    openclaw connect discord --token YOUR_DISCORD_BOT_TOKEN --guild-id YOUR_SERVER_ID

    Option C: WhatsApp

    WhatsApp connection requires the gateway to be accessible via a public URL. Use SSH tunneling:

    ngrok http 3210
    openclaw connect whatsapp --webhook-url https://your-ngrok-url.ngrok.io

    Step 4: Your First Conversations

    Once connected, let's try some basic commands through Telegram (or your preferred platform).

    Basic Assistant Tasks

    Get a daily briefing:

    You: Morning! Give me a briefing on what I have today.
    
    OpenClaw: Good morning! Here's your briefing:
    📅 Calendar: 2 meetings — standup at 10 AM, design review at 3 PM
    📧 Email: 12 unread, 3 flagged as urgent
    🌤️ Weather: Sunny, 72°F in your area
    📋 Tasks: 5 items due today, 2 overdue from yesterday

    Search the web:

    You: Find me the latest React Native release notes.
    
    OpenClaw: React Native 0.75 was released on August 12, 2024. Key changes:
    - Yoga 3.1 layout engine update
    - New Architecture is now stable
    - Improved Metro bundler performance
    - TypeScript 5.2 support

    File operations:

    You: Summarize /Documents/meeting-notes.txt
    
    OpenClaw: Summary: The Q3 planning meeting covered 3 main topics:
    1. Product roadmap priorities for mobile app rewrite
    2. Budget allocation for the new ML pipeline
    3. Hiring plan: 2 senior engineers needed by October

    Step 5: Setting Up Skills

    Skills are the superpower of OpenClaw — repeatable workflows defined in Markdown or TypeScript.

    Installing a Community Skill

    openclaw skills list
    openclaw skills install github-pr-reviewer
    openclaw skills install daily-standup-reporter
    openclaw skills install expense-tracker

    Writing Your Own Skill

    Create a new file at ~/.openclaw/skills/morning-briefing.md:

    ---
    name: morning-briefing
    description: Sends a personalized morning briefing
    triggers:
      - schedule: "0 8 * * 1-5"
      - keyword: "morning briefing"
    ---
    
    # Morning Briefing Skill
    
    ## Instructions
    1. Check Google Calendar for today's events
    2. Fetch the top 5 important unread emails
    3. Get the current weather for the user's location
    4. Summarize all open GitHub PRs assigned to the user
    5. Format and send a clean briefing via Telegram

    Reload skills to activate:

    openclaw skills reload

    Skills via TypeScript (Advanced)

    import { Skill } from '@openclaw/sdk';
    
    export const expenseTracker: Skill = {
      name: 'expense-tracker',
      description: 'Tracks and categorizes expenses from email receipts',
      
      async execute(ctx) {
        const emails = await ctx.gmail.search(
          'subject:(receipt OR invoice) newer_than:7d'
        );
        
        const expenses = [];
        for (const email of emails) {
          const parsed = await ctx.ai.extract(email.body, {
            amount: 'number',
            vendor: 'string',
            category: 'string',
            date: 'date'
          });
          expenses.push(parsed);
        }
        
        await ctx.memory.append('expenses.json', expenses);
        const total = expenses.reduce((sum, e) => sum + e.amount, 0);
        return `Found ${expenses.length} expenses totaling $${total.toFixed(2)}`;
      }
    };

    Step 6: Proactive Intelligence and Cron Jobs

    Configure heartbeats in ~/.openclaw/heartbeat.yml:

    heartbeats:
      - name: github-monitor
        schedule: "*/15 * * * *"
        task: |
          Check for new GitHub notifications. If there are any failing CI checks
          on my PRs, diagnose the issue and suggest a fix in Telegram.
    
      - name: email-triage
        schedule: "0 9,13,17 * * 1-5"
        task: |
          Check for urgent unread emails. If any require immediate action,
          summarize them and send an alert via Telegram.
    
      - name: code-backup
        schedule: "0 22 * * *"
        task: |
          Run git status in ~/Projects. If any repos have uncommitted
          changes, send a reminder message.

    Developer Workflow Automation

    You: Set up monitoring for my React Native app's test suite.
    
    OpenClaw: Done! I've set up:
    1. ✅ Webhook to receive Sentry error events
    2. ✅ Cron job to run 'yarn test' every 4 hours
    3. ✅ GitHub Actions notification listener
    4. ✅ Auto-issue creation for any failing tests
    
    When tests fail, I'll diagnose the error, suggest a fix, and optionally
    open a PR with the fix for your review.

    Step 7: Memory and Personalization

    Integrating with Obsidian

    openclaw integrate obsidian --vault-path ~/Documents/ObsidianVault

    Now OpenClaw can read and write to your vault:

    You: Create a meeting note for today's standup.
    
    OpenClaw: Created: /Meetings/2026-03-20-standup.md ✅
    Linked to your Q1 project note and tagged #standup.

    Teaching OpenClaw Your Preferences

    Tell it your preferences directly and it remembers them forever:

    You: Remember that I prefer concise responses under 3 bullet points.
         My timezone is UTC+6.
    
    OpenClaw: Saved your preferences:
    - Response style: Concise, max 3 bullet points
    - Timezone: UTC+6 (Bangladesh Standard Time)
    I'll apply these to all future interactions.

    Real-World Examples

    Example 1: Developer on a Walk

    You: I'm going for a walk. Review my open PRs while I'm gone.
    
    [30 minutes later]
    OpenClaw: Walk summary! Reviewed 3 PRs:
    - PR #47: Approved ✅ — Clean implementation, left 2 minor suggestions
    - PR #48: Changes requested ⚠️ — Missing error handling in fetchUser()
    - PR #49: CI failing ❌ — Ran snapshots update, opened fix PR #50

    Example 2: Personal Finance

    You: How much did I spend on food this month?
    
    OpenClaw: Food spending (March 2026):
    - Restaurants: $342.50 (18 transactions)
    - Groceries: $218.75 (6 transactions)
    - Coffee shops: $67.20 (14 transactions)
    Total: $628.45
    
    Tip: 23% over your usual average of ~$510/month.
    Want me to set up weekly budget alerts?

    Example 3: Research Assistant

    You: Compare the top React Native navigation libraries.
    
    OpenClaw: Here's the comparison:
    
    Library            | Stars | Size   | TypeScript | Expo
    React Navigation  | 23k   | 180KB  | ✅         | ✅
    Expo Router       | 4.2k  | 220KB  | ✅         | ✅
    RN Router Flux    | 9.1k  | 95KB   | ✅         | ⚠️
    Wix Navigation    | 13k   | 310KB  | ✅         | ❌
    Solito            | 3.8k  | 45KB   | ✅         | ✅
    
    Recommendation: React Navigation v7 for your stack.
    Want me to scaffold a project with it configured?

    Security Best Practices

    1. Always bind the gateway to localhost:

    gateway:
      bind: localhost

    2. Use SSH tunneling for remote access:

    ssh -L 3210:localhost:3210 user@your-home-ip

    3. Enable skill sandboxing:

    sandbox:
      enabled: true
      network_access: restricted
      filesystem_access: scoped

    4. Regularly rotate API keys:

    openclaw config rotate-key --provider anthropic

    Troubleshooting Common Issues

    Gateway won't start

    docker ps
    openclaw gateway restart
    openclaw logs --tail 50

    Messages not arriving on Telegram

    openclaw connect telegram --verify
    openclaw status --verbose

    High memory usage

    openclaw memory compact
    openclaw config set memory.max_context_messages 50

    Conclusion

    OpenClaw represents a fundamental shift in how we think about AI assistants. It's not a tool you visit — it's a teammate that lives alongside you, learns your preferences, and gets things done proactively.

    Key advantages:

  • Privacy by default: Your data stays on your machine
  • Genuinely autonomous: It acts, not just answers
  • Infinitely extensible: Build skills for any workflow
  • Platform-agnostic: Works through your existing chat apps
  • Developer-friendly: CLI, TypeScript SDK, and open-source codebase
  • Start with the one-liner install, connect your Telegram bot, and see what it does in the first week. You'll never want to go back to a plain chatbot again.

    ---

    *Have questions about setting up OpenClaw or building custom skills? Drop a message below — I'd love to help!*

    Tags

    #AI#OpenClaw#Automation#Personal Assistant#Open Source#Developer Tools

    Share this article

    Let's Work Together

    Need help with your mobile app or have a project in mind?

    Sani Mridha - Senior React Native Developer | iOS & Android Expert