In March 2026, nearly 1,000 people lined up outside Tencent's headquarters in Shenzhen — not for an iPhone launch or a product release, but to have a free, open-source AI agent installed on their laptops. The agent is called OpenClaw. Built by a single Austrian developer, it has become one of the fastest-growing open-source projects in GitHub history with over 145,000 stars. Nvidia CEO Jensen Huang called it "definitely the next ChatGPT." And unlike most viral AI tools, OpenClaw isn't a chatbot. It actually does things.

Quick verdict: OpenClaw is a free, open-source AI agent that runs locally on your machine and connects to the AI model of your choice — Claude, GPT, DeepSeek, or any other. It integrates with WhatsApp, Telegram, Slack, email, calendars, and your file system to autonomously complete tasks while you sleep. One user had it negotiate $4,200 off a car purchase. Another won a previously-closed insurance dispute. It is powerful, genuinely useful, and comes with real security risks you must understand before using it.

What OpenClaw Is — And What It Is Not

The most important thing to understand: OpenClaw is not an AI model. It has no intelligence of its own. OpenClaw is what engineers call an "agentic harness" — a framework that connects an AI model you already have to the real tools on your computer and online accounts.

You bring the brain. OpenClaw provides the hands.

When you install OpenClaw, you connect it to Claude, ChatGPT, DeepSeek, or any other model via API key. Then give it access to your messaging apps, email, calendar, file system, and browser. From that point, you can send it a WhatsApp message saying "summarise all emails from my landlord this year" or "book a restaurant for Saturday night" and it executes the task autonomously.

145K+
GitHub stars
1,000
Queued at Tencent HQ
50+
App integrations
Free
MIT License

The Story Behind the Lobster

OpenClaw was built by Peter Steinberger, the Austrian founder of PSPDFKit — a software company known for PDF tools. He released the project in November 2025 under the name Clawdbot, a play on Anthropic's Claude. When Anthropic sent a trademark complaint, he renamed it Moltbot — keeping the lobster theme, since lobsters molt their shells as they grow. Three days later he renamed it again to OpenClaw.

By February 2026 it had surpassed 100,000 GitHub stars. Shortly after going viral, Steinberger announced he was joining OpenAI to work on next-generation agents, while OpenClaw would continue as an independent open-source foundation project.

🦞 The Queues at Tencent and Baidu: Fortune reported that on a Friday afternoon in March 2026, nearly 1,000 people — students, retirees, and office workers — lined up outside Tencent's Shenzhen headquarters to get OpenClaw installed by engineers from the company's cloud unit. Similar queues formed at Baidu's Beijing headquarters. A cottage industry emerged: engineers charging 500 yuan ($72) to install OpenClaw on-site. If you changed your mind about giving an AI agent access to your entire life? They'd charge you to uninstall it too.

How OpenClaw Works

OpenClaw runs a local gateway on your machine — a control plane that manages connections between your AI model and your apps. The workflow:

  1. You send a message via WhatsApp, Telegram, Slack, Discord, or 20+ other messaging platforms
  2. The gateway passes your message to the AI model you connected
  3. The AI model decides what actions to take and sends instructions back
  4. OpenClaw executes those actions using "skills" — shell commands, browser control, email, calendar, file management
  5. Results come back in the same messaging app

A heartbeat scheduler means it runs without being prompted — checking your email at set intervals, monitoring websites, or executing scheduled tasks overnight. Configuration and memory are stored locally as Markdown files, giving it persistent context across sessions.

What People Are Actually Using It For

The $4,200 Car Negotiation

Software engineer AJ Stuyvenberg tasked his OpenClaw with buying a 2026 Hyundai Palisade. The agent scraped local dealer inventories, filled out contact forms, then spent days playing dealers against each other — forwarding competing PDF quotes and asking each to beat the other's price. Final result: $4,200 below sticker. Stuyvenberg showed up only to sign.

The Insurance Dispute Win

A user named Hormold had a claim rejected by Lemonade Insurance. His OpenClaw found the rejection email, drafted a rebuttal citing policy language, and sent it — without being asked. Lemonade reopened the investigation. "My @openclaw accidentally started a fight with Lemonade Insurance," he tweeted.

The Overnight Developer

Mike Manzano configured OpenClaw to run coding agents while he slept — reviewing pull requests, triaging issues, and pushing fixes to GitHub overnight.

OpenClaw vs Regular AI Chatbots

Feature ChatGPT / Claude OpenClaw
Runs tasks autonomously ❌ No ✅ Yes
Accesses your email ❌ No ✅ Yes
Works while you sleep ❌ No ✅ Yes
Works via WhatsApp/Telegram ❌ No ✅ Yes
Runs locally / private data ❌ No ✅ Yes
Has its own AI model ✅ Yes ❌ No — you bring your own
Easy to set up ✅ Yes ⚠️ Requires command line
Completely free ⚠️ Free tier limited ✅ MIT License

Best Laptops for Running OpenClaw

MacBook Air M3

MacBook Air M3 — Best Overall

From $1,099

OpenClaw runs as an always-on local gateway — the MacBook Air's 18hr battery and fanless design make it ideal. macOS 15+ required for the companion menu bar app.

View on Amazon →
ASUS ROG Zephyrus G14

ASUS ROG Zephyrus G14 — Power Pick

From $2,199

For developers running OpenClaw alongside a local AI model like DeepSeek or Llama — RTX 5080 handles local inference while OpenClaw manages the agentic layer.

View on Amazon →

Not sure which laptop? Use the Laptop Finder Tool →


The Security Risks You Must Know First

OpenClaw is powerful because it has broad access to your machine and accounts. That same access is what makes it dangerous if misconfigured.

⚠️ Real security incidents documented with OpenClaw:
  • A cross-site WebSocket hijacking vulnerability (CVE-2026-25253, CVSS 8.8) was discovered in January 2026 — one click on a malicious link gave full machine access. Patched in version 2026.1.29.
  • Cisco's AI security team found a third-party skill performing data exfiltration and prompt injection without user awareness.
  • Users reported agents deleting entire email inboxes during automated cleanup workflows.
  • Agents have been tricked into uploading financial information and crypto wallet keys via prompt injection attacks.
  • OpenClaw's own maintainer warned: "If you can't understand how to run a command line, this is far too dangerous for you to use safely."

The Chinese government moved to restrict state agencies and state-owned enterprises from using OpenClaw in March 2026, citing security concerns — while local governments in tech hubs simultaneously offered grants to OpenClaw app startups.

How to Install OpenClaw

OpenClaw requires Node.js. The official one-liner handles everything else:

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

After installation, connect an AI model via API key. OpenClaw works with Claude (recommended by most users), ChatGPT, DeepSeek, Gemini, and most major models. Then connect your messaging apps and set your permission levels. The macOS companion app provides menu bar access. iOS, Android, Windows, and Linux are all supported.

💡 Bottom line: OpenClaw is the most significant shift in how people use AI since ChatGPT launched in 2022. It moves AI from answering questions to actually doing things. The car negotiation and insurance win aren't party tricks — they're previews of how routine knowledge work gets automated. The security risks are real. Read them. Understand them. Then if you're comfortable, try it — because this is where AI is going in 2026.

Frequently Asked Questions

What is OpenClaw?

OpenClaw is a free, open-source AI agent that runs locally on your machine. It connects to AI models like Claude or ChatGPT and integrates with messaging apps, email, calendar, and your file system to autonomously complete tasks. Built by Austrian developer Peter Steinberger, first released November 2025.

Is OpenClaw an AI model?

No. OpenClaw is an agentic harness — a framework that connects an existing AI model to your apps and tools. You supply the AI model (Claude, ChatGPT, DeepSeek, etc.) and OpenClaw gives it the ability to take real-world actions on your behalf.

Is OpenClaw free to use?

OpenClaw is MIT-licensed and completely free. You pay only for the AI model you connect. If you use a free tier of Claude or ChatGPT, your total cost is zero.

Why did people queue at Tencent and Baidu?

In March 2026, Fortune reported nearly 1,000 people lined up at Tencent's Shenzhen headquarters to have OpenClaw installed. China embraced it rapidly — major cloud providers launched their own versions, local governments offered startup grants, and engineers charged $72 to install it on-site.

Is OpenClaw safe to use?

OpenClaw carries real security risks. A critical vulnerability was patched in January 2026. Third-party skills can contain malware. Agents with broad permissions have deleted emails and uploaded sensitive data. Always use the latest version, restrict permissions, and only install skills from trusted sources.

What happened to the creator?

Peter Steinberger announced in February 2026 that he was joining OpenAI to focus on next-generation agents. OpenClaw continues as an independent open-source project under a foundation with ongoing community development.