What Is OpenClaw (Formerly Moltbot/Clawdbot)? The AI Agent Everyone's Talking About

Last updated: March 2026

If you follow tech news at all, you've probably seen this project under at least three different names. Clawdbot. Moltbot. And now, OpenClaw.

It's the same project. It just keeps changing names. Here's the full story and what it means for you as a builder.

---

The Wild Rebrand Story

Austrian developer Peter Steinberger built Clawdbot as a weekend project in November 2025. It was a simple idea: a personal AI assistant that runs on your own computer and connects to your messaging apps.

It went viral. 100,000 GitHub stars in days. 2 million visitors in the first week. Then things got complicated.

January 27, 2026: Anthropic (the company behind Claude AI) sent a trademark notice. "Clawd" was too close to "Claude." Fair enough. Steinberger renamed it to Moltbot, a nod to how lobsters molt their shells.

But during the rename, something wild happened. Scammers exploited a 10-second window between the GitHub and Twitter account name changes. They grabbed the old accounts and launched a fake crypto token ($CLAWD on Solana) that hit a $16 million market cap before people caught on.

January 30, 2026: Another rename. This time to OpenClaw. "Open" for open-source, "Claw" to keep the lobster theme. The team was more careful this time and secured everything before the switch.

February 14, 2026: Steinberger announced he was joining OpenAI to work on agentic AI. OpenClaw was transferred to an independent open-source foundation so the community could keep building.

Three names in two weeks. A crypto scam. And the creator leaving for OpenAI. You can't make this stuff up.

---

What OpenClaw Actually Does

Strip away the drama and OpenClaw is genuinely impressive tech.

It's a personal AI assistant that:

1. Runs on your own computer so your data stays with you
2. Connects to 50+ messaging platforms including WhatsApp, Telegram, Discord, Slack, and iMessage
3. Uses powerful AI models like Claude, GPT, and Gemini to understand and respond
4. Can actually do things for you like browse the web, manage files, schedule calendar entries, even send emails

The big difference from something like ChatGPT or Claude's web interface? OpenClaw works inside your messaging apps. You text it on WhatsApp and it responds. It can also be proactive. It has a built-in "Heartbeat Engine" and cron jobs that let it take action on its own schedule, not just when you ask.

With 190,000+ GitHub stars, it's one of the most popular open-source projects of 2026.

---

The Security Problems Worth Knowing About

The rapid growth came with serious security issues that are worth mentioning:

Exposed instances. Security researchers found hundreds of unprotected OpenClaw instances through Shodan (a search engine for internet-connected devices). These had leaked API keys and open command execution access. The latest version (v2026.1.29) removed the insecure "no auth" option entirely.

Malicious VS Code extension. A fake "ClawdBot Agent" extension appeared that installed remote access trojans on developers' machines. Always verify extensions come from official sources.

The crypto scam. As mentioned above, the $CLAWD token scheme during the name change exploited the confusion. Not related to the actual project, but people lost real money.

The project team has addressed these issues, but it's a good reminder that popular open-source tools can attract bad actors.

---

The Catch: It's a Personal Tool, Not a Platform

Here's the honest part. OpenClaw is powerful, but it's a personal assistant for you, the operator. It's not infrastructure for building a product on top of WhatsApp or Instagram.

To get it running, you need to:

• Install Node.js (version 22+)
• Run command-line instructions
• Configure API keys for the AI models you want to use
• Set up a "Gateway" process for your messaging platforms
• Keep your machine on and troubleshoot when things break

And even once it's running, the channel plumbing is yours to maintain: unofficial integrations, your own machine as the runtime, no signed webhooks, no production-grade delivery guarantees. That's fine for a personal assistant. It's a different story when you want to ship something on WhatsApp or Instagram for real customers.

Cost-wise, it depends on how much you use it. Light usage runs about $10-30/month in AI model costs, moderate usage $30-70/month, and heavy usage $70-150/month. You can use local models through Ollama to bring costs down, but the quality trade-off is real.

The OpenClaw docs are solid, but they assume you're comfortable with technical setup and that the thing you're building is for yourself.

---

What If You Want to Build a Real Messaging Integration?

A lot of developers don't actually want a personal lobster assistant. They want to build something on top of WhatsApp, Instagram, or Messenger: a booking flow, a lead-capture form, an order-status notifier, an AI agent that lives in a customer's DMs and writes data back to their own stack.

For that you don't want a desktop daemon. You want infrastructure: one API that abstracts all three channels, signed webhooks you can trust in production, and a way to collect structured data in-chat instead of parsing free-text messages yourself.

That's what Wabery is. It's not a packaged "AI agent" we run for you. It's the messaging API and toolkit you build your own solution on:

• A unified channels API for WhatsApp, Instagram, and Messenger, so you write your integration once.
• Signed event webhooks so every inbound message, status change, and form submission lands in your backend, verifiable and replayable.
• Native WhatsApp Flows (in-chat forms) so you can collect structured data without inventing your own parser.
• Automations for the deterministic parts (auto-replies, routing, follow-ups) that you'd otherwise reimplement on every project.
• A CLI and an MCP server so you can scaffold, test, and wire things up from your terminal or straight from an AI coding agent.

Here's how the two approaches compare:

The point isn't that one is better. OpenClaw is great at being a personal tool. Wabery is for when you want to build the solution yourself and ship it.

---

The Bigger Picture

OpenClaw matters because it shows where AI-in-messaging is heading. The idea that an AI agent lives in your chat apps, understands context, and takes action is going to become normal.

The interesting question for builders isn't "which assistant should I install" but "what do I want to build, and on what?" If you're shipping something for real users on WhatsApp or Instagram, you want infrastructure underneath you, not a daemon on your laptop.

With Wabery you get the primitives and assemble the solution yourself. A WhatsApp Flow collects the data, a signed webhook delivers it to your stack, an automation handles the routine path, and your code does whatever's specific to your product. You can stand up a working booking or lead-capture flow in an afternoon, not a sprint.

---

Quick Summary

What is OpenClaw? An open-source AI assistant that runs on your computer and connects to WhatsApp, Telegram, Discord, and other messaging platforms. It went through three names: Clawdbot (Nov 2025), Moltbot (Jan 27, 2026), and OpenClaw (Jan 30, 2026).

Who built it? Peter Steinberger, an Austrian developer who has since joined OpenAI. The project is now run by an independent open-source foundation.

Who is it for? Developers and tech enthusiasts who want a personal AI assistant they fully control and are comfortable with technical setup.

Where can I learn more? The official site is at openclaw.ai

What if I want to build a messaging integration for customers? That's what Wabery is for. One API across WhatsApp, Instagram, and Messenger, signed webhooks, native WhatsApp Flows, a CLI, and an MCP server, so you build the booking flow, lead capture, or AI agent yourself, fast.

See how fast you can ship a messaging integration when the primitives are already there.