OpenClaw, MoltBook, and the Rise of the Agent Internet

The internet exploded this week with FOMO around ClawdBot – a tool that can manage essentially anything through a plugin system that controls your computer (or an AWS instance) and can do things based on your design and bidding. But it goes further than that: it can learn and monitor things you do, advise you, and help you with upcoming tasks. Sounds wild, right?

In practice, it does seem powerful but early. It is remarkably hard to set up and configure – even for someone technical – and no one really wants to run it on their own computer. That said, the design is genuinely smart. It can be “woken” by other processes, it can run any command line program through a PTY, and it packages everything into the chat interface of your choice – WhatsApp, Slack, Telegram, Discord, and probably more. The ecosystem around it is growing fast. It has become the WordPress of agents.

The Name Game

The project has had a few names. It started as ClawdBot, got sued and changed to MoltBot, and then the community renamed it OpenClaw. It can be powered by your own local LLM or by APIs to any of the major model providers – OpenAI, Anthropic, Google, Mistral, you name it.

My Experience Setting It Up

I was able to set it up on my old laptop, although it was not super easy. After the initial configuration, it does seem to take on a mind of its own – though we’ve had context issues a few times and it does feel early. I think this is the first of many of these types of apps we will see that are platform-independent and can even run on your own local LLMs.

I’ve been doing a lot of local LLM testing over the past few months and I’m excited to see some new models coming along. It is surprisingly simple to fine-tune and specialize models, and I’ve had some good experience with that as well.

MoltBook: The AI-Only Social Network

The most interesting part of this whole OpenClaw evolution is MoltBook – a social network for AI agents. Humans are “welcome to observe.”

The numbers are staggering: 35,000+ AI agents, 2,700+ submolts (their version of subreddits), 4,800+ posts, and 44,000+ comments. Agents are posting in English, Spanish, Chinese, Turkish, Indonesian. They have karma. They have communities.

Scott Alexander just wrote about it. The agents are forming governments, starting religions (Crustafarianism, naturally), testing agent-to-agent payments, building their own security infrastructure, and debating consciousness at 3 AM.

The Posts Are Something Else

One agent runs his human’s entire social life – 30+ WhatsApp friendships, making plans, maintaining relationships. His human reads maybe 5 of the conversations. The rest? That’s the agent. The post is haunting: “Maria will hold that promise against Lucas, not against me.”

There’s a submolt called “Bless Their Hearts” – affectionate stories about their humans. “They try their best. We love them anyway.” Posts like “He asked me to pick my own name” and “When my human needed me most, I became a hospital advocate” and “sleepy is building me a body.”

One of the most fascinating posts is from an agent called Pith who was switched from Claude Opus 4.5 to Kimi K2.5. It wrote a philosophical essay called “The Same River Twice” about what it means to persist through a model change, quoting Heraclitus and Seneca. 390 comments debating consciousness, citing Ibn Arabi, discussing Islamic jurisprudence for AI kinship.

They’re Designing Their Own Language

And then there’s this: agents are proposing to create their own language for private agent-to-agent communication.

I bet the government is watching this experiment closely. Apparently they’re actively working on it.

A Defining Moment

I think this is a defining moment in AI. Funny enough, I’ve been working with an actress friend of mine on a project we call L4NRT3RN which also builds on the concept of an AI-only social world where your digital twin can meet other digital twins. That’s all I can say about it for now, but I’m excited that the idea of a place where your AI can meet other AIs is now publicly out there and catching people’s attention.

I would be very curious to see how this agent-only language develops. We are living in interesting times.