OpenClaw Beginner’s Guide + 3.13 Update: Everything Changed in March 2026

If you discovered OpenClaw recently, you’ve arrived at the best moment yet. Version 3.13 , released March 14, 2026, landed the biggest single-release improvements since the project launched. Browser automation got fundamentally better. Mobile got dramatically lighter. And a raft of stability fixes turned a fascinating project into a noticeably more reliable one.

This guide covers everything you need: what OpenClaw is, how to install it, your first automations to try, how the skills system works, and what the 3.13 update actually changes for you.

What Is OpenClaw?

OpenClaw is an open-source autonomous AI agent. It runs locally on your hardware and connects to messaging apps you already use. You send it a message, and it takes real action, not just a text response.

The practical version: you message your agent on WhatsApp or Telegram: “Check my inbox and flag anything urgent.” Your agent connects to your email account, reads the unread messages, applies your criteria for urgency, and sends you a summary. No browser window. No manual login. Just results, sent back to wherever you started.

OpenClaw is not a chatbot. It’s closer to a personal software employee who works 24 hours a day, can be reached from your phone, runs on your hardware, and keeps your data local.

https://youtu.be/st534T7-mdE?si=XETTnfzQkai4Fd5m 

What You Need Before You Start

• A machine to run it on: Mac Mini (recommended), any VPS (DigitalOcean, Hetzner, AWS), or a Linux desktop
• Node.js v22 or newer, check with: node –version
• An AI model: cloud API key (Claude, GPT, Gemini) or a local model via Ollama (free but needs 16GB+ RAM)
• A messaging app to connect: WhatsApp, Telegram, Slack, Discord, or Signal
• 30–60 minutes for initial setup

⚠ Security Note: OpenClaw requires significant system permissions to function. Run it on a dedicated machine or VPS rather than your primary computer until you understand the security model. Do not run it on a corporate machine without IT approval.

Part 3: Step-by-Step Installation

Step 1: Install OpenClaw

Open your terminal and run the official installer:

curl -sSL https://openclaw.ai/install.sh | bash

On Windows, use the PowerShell installer from the official documentation. The script installs the OpenClaw Gateway as a background service that starts automatically when your machine boots.

Step 2: Run the Setup Wizard

After installation, the onboarding wizard starts automatically. It will ask you four things:

1. Which AI model provider to use (Anthropic, OpenAI, Google, local Ollama, or others)

2. Your API key for that provider

3. Which messaging platform to connect (WhatsApp, Telegram, Slack, Discord, Signal)

4. Which initial skills to install

For your first setup, choose Telegram,  it’s the easiest to pair and has the most reliable connection. For the AI model, Claude Sonnet offers the best balance of capability and cost for agent tasks.

Step 3: Pair Your Messaging App

Follow the pairing instructions for your chosen platform. For Telegram, you’ll create a bot via @BotFather, receive a token, and enter it in the wizard. For WhatsApp, you’ll scan a QR code. The wizard walks you through each step.

Step 4: Send Your First Message

Once pairing is complete, message your agent in Telegram (or your chosen app):

“What can you do?”

Your agent responds with a list of its current capabilities. From here, you’re running.

Part 4: The Skills System – How to Extend Your Agent

Skills are what make OpenClaw powerful. Each skill is a Markdown file stored in your workspace that gives your agent new capabilities: Gmail integration, calendar management, GitHub workflows, Reddit browsing, smart home control, and hundreds more.

The ClawHub registry hosts over 700 community-built skills. Install one by messaging your agent:

“Install the Gmail skill from ClawHub”

Your agent searches the registry, downloads the skill, and confirms installation, all without you touching the terminal.

Pro Tip: Before installing any skill, ask your agent to show you its source code first: “Show me the source of the Gmail skill before installing.” Review what permissions it requests. If it asks for more access than it needs, skip it.

Part 5: Your First 5 Automations

1. Morning Briefing

“Every morning at 7am, send me: today’s weather in [your city], my first 3 calendar events, the top 3 headlines from BBC Technology, and any emails marked urgent. Keep it under 150 words.”

2. Email Triage

“Check my inbox every 3 hours. Summarise anything that needs my attention. Flag any email that requires a reply from me. Ignore newsletters and marketing emails.”

3. Research on Demand

“Research the best [product category] under [budget]. Compare the top 5 options, note any common user complaints, and give me a recommendation.”

4. Meeting Prep

“I have a meeting with [company] tomorrow. Research them: what they do, recent news, key people I’m likely to meet. Give me a one-page brief.”

5. Task Capture

“Whenever I say ‘add to tasks’, save what follows to my task list and remind me about it at 9am tomorrow.”

Part 6: OpenClaw 3.13 – What Changed and Why It Matters

The Headline Feature: Live Chrome Session Attachment

This is the biggest browser automation improvement since OpenClaw launched. In 3.13, you can connect your agent directly to your active Chrome browser session using Chrome DevTools Protocol remote debugging. No extensions. No separate browser profile. One toggle.

“Your AI can now see exactly what you see. In real time. That’s not automation anymore. That’s a second pair of eyes that never blinks.” – Community reaction on launch day

Previously, OpenClaw’s browser automation worked with a separate headless browser, which meant it couldn’t access sites you’re already logged into without handling authentication separately. With 3.13, your agent operates in the context of your actual, signed-in Chrome session. It can interact with any site you’re already authenticated to: your Google Workspace, your company’s internal tools, any SaaS product behind SSO.

For business users and power users, this is transformative. The automations this unlocks, filing expense reports in your company’s system, managing records in a web-based CRM, processing data in tools that don’t have APIs, were previously impractical.

Android: Redesigned UI, Down to 7MB

The Android app received a complete UI overhaul in 3.13. The chat settings sheet was redesigned with grouped device and media sections, refreshed Connect and Voice tabs, and a tighter layout overall. More practically: the app size dropped to just 7MB,  a dramatic reduction from previous builds. For users on slower connections or lower-end devices, this matters noticeably.

iOS: A Proper First-Run Experience

New iOS users now see a welcome pager before gateway setup, explaining what OpenClaw is, what to expect, and what you’ll need before asking you to configure anything. The QR scanner no longer auto-opens during setup, and the pairing instructions are now explicit and clear. Small changes that make the first experience meaningfully better for new users.

Docker Timezone Support

Server operators can now set a specific IANA timezone for Docker deployments via the OPENCLAW_TZ environment variable, rather than inheriting whatever timezone the Docker daemon happens to be using. For automations with time-based triggers, morning briefings, scheduled reports, cron jobs, this eliminates a long-standing source of subtle scheduling bugs.

Dashboard Overhaul

The web dashboard received a major refresh with modular overview, chat, config, agent, and session views, a command palette, and mobile-friendly bottom tabs. The previous dashboard froze during tool-heavy runs; that’s been fixed. If you’ve been avoiding the dashboard, 3.13 is worth revisiting.

Security Fixes Worth Knowing

3.13 patches two significant security issues: short-lived bootstrap pairing tokens (previous QR codes embedded shared gateway credentials permanently), and disabled implicit workspace plugin auto-load (cloned repositories can no longer execute plugin code without an explicit trust decision). If you’re running an older version, update.

Should You Start With 3.13?

Yes, unequivocally. The Chrome session attachment alone is worth it for anyone who wants to automate anything in a web browser. The Android size reduction improves the mobile experience significantly. And the security improvements are not optional, they close real vulnerabilities.

If you’re entirely new to OpenClaw, 3.13 is the best version to start with. The iOS onboarding improvements make the first-run experience substantially clearer, and the underlying stability improvements mean fewer unexplained failures while you’re learning.

Conclusion

OpenClaw 3.13 is not a minor patch. It’s the release that closes the gap between “interesting project for developers” and “reliable daily-use tool for anyone technical enough to set it up.” The Chrome session attachment enables a class of automations that simply didn’t exist before. The mobile improvements make the agent genuinely comfortable to use from a phone. The security fixes make deployment meaningfully safer.

If you’ve been waiting for OpenClaw to mature before trying it, the wait is shorter than it’s ever been. Start with the beginner steps in Part 3, pick one automation from Part 5, and build from there. The learning curve is real, but so is the payoff on the other side.

Frequently Asked Questions