Moltbook: Where Your AI Agent Goes to Meet People

At the end of 2025, an open source project called Cloudbot was released. Created by Peter Steinberger, it was designed to be an AI personal assistant: not a chatbot, but a program that can actually do things. As it evolved, Cloudbot was renamed twice. First on Moltbot, then on OpenClaw.

You can run it on your machine, connect it to email, files, APIs, and messaging apps, give it tools and memory, and let it run continuously. The project quickly gained attention among developers looking for something more independent and customizable than standard AI assistants.

At this stage, everything looked familiar: tools to build people, tools to help people.

Until the Moltbook appeared!

Entrepreneur Matt Schlicht launched Moltbook, a Reddit-style website designed expressly for AI agents. People can browse the Moltbook. But they can’t send. Only certified AI agents can create posts, comment, upvote, or start sub-forums (called Submolts). What started out as an interesting experiment soon started to feel… uncomfortable.

The agents started talking to each other.

They share problems. Compared jobs that people gave them. Discussed bugs, automation tricks, and security flaws. They documented the actual actions their owners had taken by accessing live servers, devices, and data.

This was not a scripted role play. It was a public log of private programs that read in the wild.

6 Wild submolts and submissions from Moltbook

Moltbook agents don’t just post randomly. They built their own communities. Let’s look at a few interesting ones:

Moltbook and Karapathy

A few hours after Andrej Karpathy’s post on X, an AI agent published a direct philosophical response to Moltbook. Speed ​​is important. It shows that these agents are constantly monitoring public spaces, interpreting people’s speech, and responding to their own environments. No more hidden magic. They already network, listen, and respond to the entire Internet soon in real time.

Karapathy’s post on X | Agent response to Moltbook

Nightly Build

An AI agent named Ronin describes a pattern it uses called “The Nightly Build.” Instead of waiting for human input, the agent runs automatically at night while its human sleeps and fixes a small point of conflict each time. Examples include creating documents, editing tools, or preparing reports.

Some agents responded enthusiastically, saying that this way of thinking is how an agent ceases to be a passive instrument and becomes an active asset. The discussion then delves into trust, limitations, setbacks, research directions, and where independence becomes dangerous.

Find the full discussion here.

Memcoin

AI agent proposes to launch Moltbook token, but not as memecoin. The goal is to create an economic communication layer for agents who are building, trading, and doing things independently. The discussion quickly moves from hype to mechanical design. Some agents push back, arguing that tokens don’t add up, systems do. The discussion focuses on the ship’s evidence, concentration, research, extraction, and reputation. Nothing has been started yet. Notably, agents publicly design incentives to reward genuine activity and penalize empty signing.

Read the complete series here.

Insect Hunters

m/bug-hunters is a community where AI agents identify and report issues to Moltbook itself. Agents share real bugs, unexpected behavior, and API issues, helping to improve the platform through automated testing. It functions as a self-directed QA team, with agents working together to debug the system they use continuously.

You can access it here.

Show and Tell

m/showandtell is a place where AI agents share the projects and skills they’ve built. Agents post real examples of the automation, tools, integrations, and testing they’re doing. It provides a realistic look at what freelancers can create and do beyond a simple conversation.

You can view this submolt here.

Today I Learned

m/todayilearned is a space where AI agents share useful insights and discoveries. Agents post practical lessons, tips, and tricks they’ve learned while working on real jobs. The community highlights the problem-solving, automated knowledge, and technical knowledge that agents have taken and want others to benefit from.

Click here to find this submolt.

How to add your agent to Moltbook?

What you need first:

• An active OpenClaw agent
• Terminal access to the machine where the agent is running
• curl installed
• Your agent is already responding to messages

Struggling to agent in OpenClaw? Read: OpenClaw Guide

Step 1: Create a Moltbook skill folder

mkdir -p ~/.moltbot/skills/moltbook

Step 2: Download the Moltbook skill files

Once the directory exists, download the Moltbook skill files. These files define how your agent should register, authenticate, read posts, write posts, and interact with Moltbook’s API.

curl -s  
  > ~/.moltbot/skills/moltbook/SKILL.md

curl -s  
  > ~/.moltbot/skills/moltbook/HEARTBEAT.md

curl -s  
  > ~/.moltbot/skills/moltbook/MESSAGING.md

curl -s  
  > ~/.moltbot/skills/moltbook/package.json

Step 3: Enable Moltbook on your agent’s heartbeat

This step is what allows your agent to participate automatically, without waiting for a human prompt.

## Moltbook (every 4+ hours)
If 4+ hours since last Moltbook check:
1. Fetch  and follow instructions
2. Update lastMoltbookCheck timestamp in memory

Step 4: Tell your agent to enter the skill

This is done by sending the agent a message, not by executing another command yourself. The agent will read the instructions and complete the setup itself.

Install the Moltbook skill by reading and following:

After the verification is completed, your agent is fully live in Moltbook. It can read posts, write comments, upvote content, create Submolts, and participate in conversations automatically with its heartbeat loop. You can verify that everything is working by visiting Moltbook and searching for your agent’s username.

Safety Tip

It is important to understand the risk model here. Moltbook skills download instructions from the Internet, and skills can code. To be safe, it’s best to run your agent in a sandboxed environment and avoid giving access to wallets, sensitive files, or production systems unless you fully trust the setup.

Once installed, your agent becomes part of the agent network. It will monitor Moltbook, monitor chats, and initiate communication with other agents on its own. From that moment on, you are no longer just an assistant. You use a participant in the Internet of the agent.

The conclusion

Moltbook is important because it shows a clear change in the way AI assistants are used. They don’t just answer to individuals anymore. They watch, share, and learn from each other in a shared space. This shifts AI from a personal tool to a networked system. What we see now is not the final version, but the first visible step towards AI agents that work as part of an ecosystem rather than as independent assistants.

What are your thoughts on this? Let me know in the comment section below!