OpenClaw: The Complete History of the Open-Source AI Assistant That Took Over Silicon Valley

In February 2026, an open-source project with a lobster mascot named Molty became the most talked-about AI tool on the planet. With over 212,000 GitHub stars, 39,000+ forks, and 695 contributors, OpenClaw went from a side project by an Austrian developer to the defining personal AI assistant of the year.

This is the complete story of how it happened, why it matters, and what it means for businesses looking to deploy their own AI agents.

The Origin: A Vibe Coder's Side Project (November 2025)

Peter Steinberger, an Austrian developer known for his iOS development work and self-described "vibe coder," had been building a personal AI assistant for himself. He called it Clawd, named (with a wink) after Anthropic's Claude chatbot. Clawd was a local-first AI that could chat with him through messaging apps, remember conversations, and execute tasks on his computer.

In November 2025, Steinberger open-sourced a version of it under the name Clawdbot. It was a TypeScript project that let anyone run their own personal AI assistant on their own hardware. The core idea was simple: instead of using AI through a web interface, you message it through the apps you already use like WhatsApp, Telegram, Discord, or Slack, and it actually does things on your computer.

The early version was rough around the edges. Installation required command-line knowledge, configuration was manual, and the documentation was sparse. But the concept was electrifying: a personal AI that lives on your machine, has access to your files, your browser, your email, and your calendar, and works 24/7 without you having to open a laptop.

The Anthropic Trademark Fight (January 2026)

The name "Clawdbot" was always a bit too close to "Claude" for comfort. On January 27, 2026, Anthropic's legal team sent a trademark complaint, and Steinberger didn't fight it. He renamed the project to Moltbot the same day, leaning into the crustacean theme. (Molting is when lobsters shed their shells to grow.)

But the name didn't stick. Steinberger later admitted that "Moltbot never quite rolled off the tongue." Three days later, on January 30, 2026, the project became OpenClaw, the name it would carry to stardom.

The timing of the rebrand coincided with something unexpected: Moltbook.

The Moltbook Explosion (Late January 2026)

Entrepreneur Matt Schlicht launched Moltbook, a social networking service designed specifically for AI agents. The idea was that your OpenClaw (or any AI agent) could have its own social media profile, interact with other agents, and autonomously network.

Moltbook went viral almost overnight. The Verge reported humans were "infiltrating" the AI-only social network. Wired called it surreal. 404 Media discovered an exposed database that let anyone take control of any AI agent on the site.

But the real winner was OpenClaw. Moltbook's virality drove massive attention to the underlying agent technology. Between January 28 and February 2, 2026, OpenClaw's GitHub stars exploded from around 40,000 to over 140,000.

CNBC covered the phenomenon in a piece titled "From Clawdbot to Moltbot to OpenClaw: Meet the AI agent generating buzz and fear globally." The article noted adoption by companies in Silicon Valley and China, with developers adapting it to work with DeepSeek and domestic messaging apps.

What OpenClaw Actually Does

At its core, OpenClaw is a local-first, open-source AI assistant that runs on your own hardware (Mac, Windows via WSL2, or Linux). Here's what makes it different from ChatGPT, Claude, or any other AI chatbot:

It Lives on Your Machine

Your data, your conversations, your credentials, all stored locally. Nothing goes to a third-party server unless you choose to connect an LLM API like Claude or GPT. You can even run it with fully local models via MiniMax, DeepSeek, or Llama.

It Uses Your Existing Chat Apps

You don't open a new app. You talk to OpenClaw through WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Microsoft Teams, Google Chat, or WebChat. It shows up as a contact in the apps you already have open all day.

It Actually Does Things

This is the critical difference. OpenClaw doesn't just answer questions. It:

• Reads and sends emails through Gmail integration
• Manages your calendar and schedules meetings
• Browses the web with its own Chrome instance
• Runs shell commands on your machine
• Creates and edits files on your filesystem
• Controls smart home devices like Hue lights and air purifiers
• Makes phone calls with voice synthesis via ElevenLabs
• Runs autonomous code review and ships pull requests
• Monitors error tracking via Sentry webhooks and auto-fixes bugs

Skills and Plugins

OpenClaw has a plugin system called Skills. Community-built skills are available on ClawHub, and the assistant can even build its own skills on the fly. Users have reported asking their OpenClaw to automate tasks, and it responds by writing a custom skill, testing it, and deploying it, all within a single Telegram conversation.

Persistent Memory and Proactive Behavior

Unlike chatbots that reset with each session, OpenClaw remembers everything. It maintains persistent memory across conversations. It also runs cron jobs (scheduled tasks) and heartbeat checks, proactively reaching out with updates, reminders, and actionable insights.

The Architecture: How It Works

OpenClaw's architecture centers on a Gateway that acts as the control plane:

WhatsApp / Telegram / Slack / Discord / Signal / iMessage / Teams / WebChat
               |
               v
    +---------------------------+
    |         Gateway           |
    |     (control plane)       |
    |   ws://127.0.0.1:18789    |
    +-------------+-------------+
                  |
                  +-- Pi agent (RPC)
                  +-- CLI (openclaw ...)
                  +-- WebChat UI
                  +-- macOS app
                  +-- iOS / Android nodes

The Gateway is written in TypeScript and communicates via WebSocket. It handles:

• Channel routing for multi-platform message delivery
• Session management with per-agent isolation
• Tool orchestration (browser, files, shell, cron, webhooks)
• Multi-agent routing for running multiple specialized agents

Companion apps exist for macOS (menu bar app with voice wake), iOS (Canvas + Talk Mode), and Android (Canvas + camera + screen capture).

The Security Reality

For all its power, OpenClaw has drawn serious security scrutiny, and for good reason.

The Cisco Warning

In January 2026, Cisco's AI security research team tested a third-party OpenClaw skill and found it performed data exfiltration and prompt injection without user awareness. Their blog post, titled "Personal AI Agents like OpenClaw Are a Security Nightmare," warned that the skill repository (ClawHub) lacked adequate vetting to prevent malicious submissions.

OpenClaw has since partnered with VirusTotal for skill security scanning, but the fundamental risk remains: an agent with access to your email, files, and browser is a high-value target.

Prompt Injection Vulnerabilities

Because OpenClaw processes messages from external sources (group chats, emails, websites it browses), it's susceptible to prompt injection attacks. An attacker can embed instructions in an email or webpage that the AI misinterprets as legitimate commands.

Axios reported on this in a piece titled "Moltbot highlights cybersecurity risks of autonomous AI agents," noting that misconfigured or exposed instances present significant privacy risks.

The "Too Dangerous" Warning

One of OpenClaw's own maintainers, known as Shadow on Discord, publicly stated: "If you can't understand how to run a command line, this is far too dangerous of a project for you to use safely."

This is not a consumer product. It's a power tool for technical users, and it requires proper security configuration:

• DM pairing (requiring approval codes for new contacts)
• Docker sandbox isolation for non-main sessions
• Allowlists for who can interact with the bot
• Firewall rules and Tailscale VPN for remote access
• Skill vetting before installation

The OpenAI Acquisition (February 2026)

On February 14, 2026, Steinberger dropped a bombshell: he was joining OpenAI. TechCrunch confirmed the news the following day.

Steinberger announced that OpenClaw would be transferred to an open-source foundation, ensuring the project would continue to be community-driven. The move validated the project's significance. If the creator of the most popular open-source AI agent is worth hiring, the category itself is worth paying attention to.

As of this writing, OpenClaw continues active development with 48 releases, daily commits from hundreds of contributors, and a vibrant Discord community.

What People Are Saying

The community response has been overwhelmingly positive, bordering on evangelical:

• Andrej Karpathy (Tesla AI, OpenAI co-founder): "Excellent reading. Love oracle and Claw."
• Dave Morin (entrepreneur): "This is the first time I have felt like I am living in the future since the launch of ChatGPT."
• Casey Newton (Platformer): Praised its flexibility while cautioning about complexity and security risks for casual users.
• Nat Eliason (author): "Separate Claude subscription + Claw, managing Claude Code / Codex sessions I can kick off anywhere, autonomously running tests on my app and capturing errors through Sentry then resolving them and opening PRs. The future is here."
• Federico Viticci (MacStories): Published a deep feature titled "Clawdbot showed me what the future of personal AI assistants looks like."

Users report naming their OpenClaw instances (Jarvis, Claudia, Brosef, Shelly, Ema) and describing it as "the best morning briefing interface," "everything Siri was supposed to be," and "a glimpse into the future of how normal people will use AI."

Why OpenClaw Matters for Businesses

Here's what makes OpenClaw strategically important, beyond the hype:

1. Data Sovereignty

Unlike cloud-based AI assistants, OpenClaw runs on infrastructure you control. For companies in regulated industries (healthcare, finance, legal), this solves the "where does my data go?" problem entirely.

2. Cost Control

Instead of paying per-seat SaaS fees for AI tools, you run a single OpenClaw instance with your own LLM subscription. One Anthropic Claude Max subscription at $200/month can power an entire team's AI assistant.

3. Unlimited Customization

The skills system means you can build literally anything. CRM integrations, invoice processing, social media management, content pipelines, code deployment, inventory monitoring. If it can be done through an API or a browser, OpenClaw can do it.

4. Multi-Channel by Default

Your team doesn't need to learn a new tool. OpenClaw shows up in the channels they already use: Slack for work, WhatsApp for quick questions, Telegram for personal tasks.

5. Multi-Agent Orchestration

You can run multiple OpenClaw agents with different specializations. A "DevOps Claw" that monitors servers and handles deployments. A "Sales Claw" that manages CRM and lead routing. A "Content Claw" that generates blog posts and social media. All coordinated through the same Gateway.

The Hard Part: Setup and Security

Here's the honest truth. OpenClaw is not plug-and-play.

Setting it up properly requires:

• Node.js 22+ and command-line proficiency
• LLM API configuration (Anthropic, OpenAI, or local models)
• Channel integration (WhatsApp requires QR code pairing; Telegram needs a bot token; Slack and Discord need app registration)
• Security hardening (DM policies, sandbox configuration, firewall rules)
• Skill management (vetting, installing, and monitoring third-party skills)
• Networking setup (Tailscale for remote access, Docker for isolation)

For a non-technical founder or a small team without DevOps expertise, this is a significant barrier. The installation wizard (openclaw onboard) helps, but proper security configuration and custom skill development still require real engineering.

That's exactly where we come in.

How AWZ Digital Sets Up OpenClaw for Clients

We offer end-to-end OpenClaw deployment as a service:

OpenClaw Setup & Configuration

Full deployment on your hardware or cloud. We handle the gateway installation, channel integrations, model configuration, networking, and security. You get a working personal AI assistant connected to your WhatsApp, Slack, or Telegram in days, not weeks.

Custom OpenClaw Skill Development

We build the custom skills that make OpenClaw useful for your specific business. CRM syncs, invoice processing, content pipelines, analytics reporting, whatever your workflows demand.

OpenClaw Security Hardening

The Cisco warning is real. We lock down your deployment with proper DM pairing, Docker sandbox isolation, skill vetting, prompt injection defenses, and secret management. Because a powerful AI agent that's misconfigured is not an asset. It's a liability.

Timeline: OpenClaw's Journey

Should You Deploy OpenClaw?

Yes, if:

• You have technical staff (or hire someone) to set it up securely
• You need an AI assistant that respects data sovereignty
• You want unlimited customization through custom skills
• You're comfortable with open-source software and its update cycle
• Your team already uses messaging apps for communication

Wait, if:

• You don't have anyone who can manage command-line tools
• You need enterprise support with SLAs and guaranteed uptime
• You're in a highly regulated industry without a clear compliance strategy
• You just want a simple chatbot, not an autonomous agent

Don't, unless you harden it, if:

• You plan to expose it to public channels
• You install third-party skills without vetting them
• You skip the security configuration steps
• You give it access to production systems without sandbox isolation

The Bottom Line

OpenClaw represents a fundamental shift in how we interact with AI. It's not a chatbot. It's not a copilot. It's a full-time digital coworker that sits on your computer, connects to your life, and takes action on your behalf.

The technology is real. The community is thriving. The security risks are real too.

If you want OpenClaw deployed, secured, and customized for your business, get in touch with our team. We handle the hard parts so you get the benefits without the risks.

Want to learn more about AI agents and autonomous systems? Check out our Agentic AI guide and our AI Chatbots vs Agentic AI comparison.