Introducing OpenClaw

From Clawd to OpenClaw

What started as a personal experiment in November 2025 has grown into something none of us expected. The idea was deceptively simple: what if you could have a truly personal AI assistant that runs on your own machine and connects to the messaging platforms you already use?

I had been frustrated with the walled-garden approach of existing AI assistants. They lived in their own apps, couldn’t talk to my existing tools, and sent all my data to someone else’s servers. So over a long weekend, I wired up a local AI gateway to WhatsApp using a headless client, gave it a personality, and called it “Clawdbot.” The first version was rough — about 400 lines of TypeScript, no plugin system, no configuration UI. But it worked. I could message my AI assistant from the same app I used to message my friends and family.

I pushed it to GitHub on a whim, shared it in a couple of Discord servers, and went to bed. By morning, the repository had 2,000 stars. By the end of the week, it had 30,000. Developers were forking it, adding Telegram support, writing custom skills, and deploying it on everything from Raspberry Pis to enterprise Kubernetes clusters. The community had spoken: people wanted this.

The Name Changes

Rapid growth brought unexpected challenges. On January 27, 2026, we received a trademark notice that made it clear we needed a new name. We briefly renamed the project to “Moltbot” — a nod to how lobsters molt their shells to grow. But the community wasn’t sold on it, and frankly, neither were we.

Two days later, on January 29, we settled on “OpenClaw.” The name captures everything the project stands for: it’s open source, it has claws (our lobster mascot has become iconic in the community), and it suggests reaching out and grabbing what you need. The rename was seamless — the community rallied around it immediately, and within 48 hours every fork, deployment guide, and tutorial had been updated.

Architecture and Design Philosophy

OpenClaw is built around three core principles: privacy first, extensibility everywhere, and deploy anywhere.

At the heart of the system is the Gateway — a lightweight orchestration layer that sits between your AI provider and your messaging channels. The Gateway manages conversation context, routes messages to the right skills, handles rate limiting, and coordinates multi-step workflows. It’s designed to be stateless where possible, so you can run multiple instances behind a load balancer without worrying about session stickiness.

Skills are the extensibility layer. Every capability in OpenClaw — from sending emails to querying databases to controlling smart home devices — is implemented as a skill. Skills are just TypeScript (or Python, or Go) modules that export a simple interface: a name, a description, a set of triggers, and a handler function. This means anyone who can write a basic function can extend OpenClaw.

The privacy-first design means your data never leaves your infrastructure unless you explicitly tell it to. Conversations are stored locally, skills execute on your machine, and the only external calls are to your chosen AI provider. You can even run OpenClaw with a fully local model if you want zero data leaving your network.

The Skills Ecosystem

The skills ecosystem has been the biggest surprise. When we launched, we included about 20 built-in skills — things like weather, reminders, and web search. Today, ClawHub (our skill registry) hosts over 5,705 skills, with more than 3,002 contributed by community members.

Skills range from the practical (CRM integrations, expense tracking, calendar management) to the creative (a skill that generates bedtime stories for kids, one that manages sourdough starter feeding schedules, another that acts as a dungeon master for text-based RPGs). The ecosystem has developed its own culture — there are “skill jams” where community members compete to build the most useful or creative skill in a weekend.

Security is paramount in a system that runs arbitrary community code. That’s why we partnered with VirusTotal to scan every skill uploaded to ClawHub. Every submission goes through automated static analysis, dependency auditing, and behavioral sandboxing before it’s published. The community also operates a peer review system where experienced contributors flag and audit popular skills.

Deploy Anywhere

One of our earliest design decisions was to avoid locking users into any particular hosting platform, and the community has taken this to heart. OpenClaw runs on:

• Railway — One-click deploy with our official template. The most popular option for getting started quickly.
• Docker — A single docker-compose up gets you a full instance with persistent storage and automatic updates.
• DigitalOcean — Community-maintained droplet images and App Platform templates.
• Self-hosted bare metal — Many users run OpenClaw on home servers, NAS devices, and even Raspberry Pis.
• Kubernetes — Helm charts for production deployments with auto-scaling and high availability.

The deployment flexibility has led to some creative setups. We’ve seen OpenClaw running on boats (for marine weather and navigation assistance), in schools (as a homework helper with parental controls), and even on a satellite ground station (don’t ask — but the skill works).

Where We Are Today

The numbers tell part of the story:

• 195,000+ GitHub stars and growing daily
• 5,705+ skills on the ClawHub registry
• 3,002+ community-built skills contributed by users worldwide
• 12,000+ Discord members across 40+ language-specific channels
• Active deployments on Railway, DigitalOcean, Docker, and more
• A thriving ecosystem of variant frameworks in TypeScript, Python, Go, Rust, and C

But numbers only capture so much. What makes OpenClaw special is the community. We have contributors from over 80 countries, regular community calls with 200+ attendees, meetup groups in 15 cities, and a documentation team that translates guides into 9 languages. People are not just using OpenClaw — they’re building businesses around it, teaching courses with it, and making it part of their daily lives.

What’s Next

We’re focused on three areas for the months ahead:

1. Gateway improvements — Better session management, multi-tenant support, and improved orchestration for complex multi-step workflows. We’re also working on a visual workflow builder that will let non-developers create sophisticated automation chains.
2. Skills ecosystem — Making it even easier to discover, install, and create skills. We’re building a new skill scaffolding CLI, improving the ClawHub search and ranking algorithms, and expanding the VirusTotal partnership to include runtime behavioral monitoring.
3. Community growth — More tutorials, better documentation, and expanding our meetup program to new cities. We’re also launching an ambassador program to recognize and support the community members who make OpenClaw great.

OpenClaw is and will always be open source under the MIT license. Your data stays on your machine, your skills live in your workspace, and the community drives the roadmap.

We’re just getting started.