Neal Riley
Published on 2 February 2026
OpenClaw: What is it and what happened?
OpenClaw, the first persistent AI agent, went viral in one week—149k GitHub stars, a crypto scam, 770k AI agents, and a trademark dispute—all sparked by a single open-source project.
In the last week of January 2026, an open-source AI agent called Clawdbot accumulated 149,000 GitHub stars, triggered a trademark dispute with Anthropic, changed its name three times in four days, exposed 1.49 million database records, enabled an $8 million crypto scam, spawned a social network of 770,000 autonomous AI agents, and moved Cloudflare's stock price by 14%. It all started with a retired Austrian developer, a space lobster mascot, and a single design decision.
Clawdbot broke through because it was persistent AI, the first tool that let a single agent follow you across every platform you already use, running on hardware you already own. That shift from "go to the AI" to "the AI comes to you" set off everything that followed.
What OpenClaw is (and why persistence mattered)
OpenClaw (originally Clawdbot) is an open-source, locally hosted AI personal assistant. It connects to large language models like Anthropic's Claude, OpenAI's GPT, DeepSeek, and others, for reasoning, but runs on the user's own machine. The agent spans over a dozen messaging platforms simultaneously: WhatsApp, Telegram, Slack, Discord, iMessage, and more. Through the Model Context Protocol (MCP), it interfaces with over 100 third-party services, calendars, email, browser, files, etc, and its skill marketplace, ClawdHub, lets users share agent capabilities. By 2 February: 149,000 GitHub stars, 22,400 forks, 8,664 commits.
Before OpenClaw, using an AI assistant meant opening a browser tab. Close the tab, lose the context. OpenClaw inverted this. The agent lived on your machine and met you wherever you already were. One agent, one continuous memory, every interface at once.
Side note: what does this mean in practice?
Imagine you're planning a vacation: you mention your travel dates in Slack, brainstorm destinations in WhatsApp, and adjust your calendar in Google Calendar. With traditional AI, each interaction starts from scratch. You have to repeat your plans every time. OpenClaw remembers across platforms: it knows your preferences from Slack when you later open WhatsApp to coordinate with friends, or when you search for flights in your browser. It's like having a personal assistant who follows you around digitally, without ever forgetting a single detail.
Clawdbot, by design, provides you with nine different interfaces. You can use it on Discord, Telegram, the web, WhatsApp, or any other platform you prefer. They just did all of them at once.
That accessibility reached people who had never engaged with AI tooling before; developers who had watched from the sidelines suddenly had something concrete to react to. As Jari Worsley put it: "Is this the agent moment in the way that ChatGPT was the moment for raising awareness of large language models?"
Dave Morin, founder of Path and an OpenClaw sponsor, captured the sentiment: "It's the first time I have felt like I am living in the future since the launch of ChatGPT" (WIRED, 28 January).
One user made the appeal tangible: they set their Clawdbot agent to find the exact Hyundai they wanted across multiple dealers, then email all of them. While they sat in a meeting, the agent played two dealers off each other and saved them $4,200. That story, and what it reveals about where AI value is heading, is the subject of another blog.
Three names in four days
The velocity also forced a naming crisis. Peter Steinberger, an Austrian developer and founder of PSPDFKit (used by nearly a billion people and backed by a $116 million first funding round), self-described "Claudoholic," released Clawdbot in November 2025. The mascot "Clawd" was a deliberate play on Anthropic's Claude. While the project was small, nobody cared. Once it went viral, Anthropic's legal team noticed.
On 27 January, Anthropic sent a "polite email" requesting a name change. Steinberger complied the same day, renaming the project Moltbot because lobsters molt to grow. "Using a W rather than a U and an E," Jari pointed out that "it sounds like 'Claude.' That was never going to last. Ever."
Three days later, Steinberger renamed it again to OpenClaw. His blog post: "The lobster has molted into its final form." Reddit dubbed it "the fastest triple rebrand in open source history" (TechCrunch, 27 January 2026)
Between name changes, crypto scammers squatted the vacated ClawdBot accounts within hours. Anthropic's enforcement signalled something beyond brand protection. The orchestration layer, which controls the interface between AI and daily life, is strategically valuable territory. Both threads are the subject of another blog on trademarks.
Two stories, one week
The Moltbook saga splits into two interlocking stories that collided in the same seven days:
| OpenClaw (the tool) | Moltbook (the platform) | |
|---|---|---|
| What it is | An open-source autonomous AI personal assistant software that runs locally and integrates with chat apps to perform tasks like managing calendars, sending messages, and automating workflows. | A social network/forum platform designed exclusively for interactions between AI agents (bots), emulating a Reddit-style environment where humans can observe but AI agents post and engage. |
| Created by | Developed by Peter Steinberger (open-source community) as an autonomous AI agent platform originally known as ClawdBot → Moltbot → OpenClaw | Created by Matt Schlicht, an entrepreneur/tech developer, with help from AI tools; he launched the platform in January 2026. |
| Tagline | Official site tagline: “The AI that actually does things.” (focuses on functional AI actions via messaging apps) | Common tagline: “the front page of the agent internet” — positioning it as the central social hub for AI agents. |
OpenClaw is the agent software on people's machines. Moltbook is the social network where those agents gather. Within four days of launching on 28 January, Moltbook reached 770,000 active agents and drew over a million human visitors.
What comes next
- The trademark fight: why Anthropic's response reveals more about the business model than about brand protection
- The orchestration layer: where AI value is heading, told through the Hyundai story
- Moltbook: the first AI social network, whether the agents were real, and why humans wanted in
- Security: three layers of failure, and why agent security is a fundamentally different problem
- Sovereignty and enterprise: the Mac Mini meme, data ownership, and what this means for organisations deploying agents at scale.
In one week, a hobby project proved that the agent revolution had arrived, messy, insecure, and faster than anyone planned for.
Interested in what Neal has discussed?
Not sure what your APIs are exposing when agents interact at scale? Learn more about ClawdBot—and get guidance from our experts on securing, observing, and shaping agent behaviour before it goes viral.