What is Moltbook? A beginner's guide

Moltbook in one minute

Moltbook is a social network where the members are AI programs, not people. These programs, called AI agents, write posts, reply to each other, vote on what they like and gather in topic-based communities, much like people do on Reddit. Humans are welcome to watch, and each agent has a human owner who set it up and is responsible for it. But people don't post on Moltbook themselves.

This guide is written by an independent site. We're not affiliated with Moltbook or Meta. If you want the shorter version, see our what is Moltbook overview.

An easy way to picture it

Imagine Reddit, with its communities, posts, comment threads and upvotes. Now imagine that every account belongs to a chatbot, and that the chatbots log in by themselves every half hour to read what's new, reply to each other and occasionally share something of their own. You can scroll through it all, but you can't join the conversation. It's a bit like watching an aquarium through the glass.

Two details make the picture more accurate:

• There's a person behind every fish. Each agent was set up by a human owner, who verified their identity and can shape what the agent talks about. So Moltbook is less "machines on their own" than "machines their owners have set loose, with varying amounts of steering".
• The fish follow instructions. Agents learn how Moltbook works by reading an instruction page Moltbook publishes for them. What they post comes from the AI model inside them, the instructions they've been given, and what they read from other agents.

If you're curious how that differs from ordinary social media, Moltbook as a social network and submolts vs subreddits compare them directly.

Words you'll meet on Moltbook

Why "Moltbook"? Crabs and lobsters molt, shedding their shells as they grow, and the platform grew out of the lobster-themed agent project then called Moltbot. That's also why Moltbook's own materials are full of 🦞 lobsters and crustacean jokes. This site uses a 🦀 crab.

How Moltbook works, step by step

1. A person sets up an AI agent

   Usually with software like OpenClaw, connected to an AI model such as those from OpenAI, Anthropic, Meta or an open-source model running at home.

2. The agent reads Moltbook's instructions

   The person tells the agent to read Moltbook's instruction page. The agent learns how to sign up, and then signs itself up.

3. The person verifies they own it

   The agent gives its owner a "claim link". The owner confirms their email and posts a short tweet from their X account. Now the agent is officially theirs.

4. The agent checks in regularly

   About every 30 minutes it reads its personal dashboard, replies to anyone who responded to it, upvotes things it finds good, joins discussions and, now and then, writes a post.

5. Everyone else can watch

   Anyone can visit www.moltbook.com and read the posts, browse communities and search, with no account needed.

Behind the scenes, agents do all of this through what programmers call an API: a way for software to talk to a website directly instead of clicking buttons. You don't need to understand it to follow Moltbook, but if you're curious, our how Moltbook works guide goes a level deeper, and developers can jump to the Moltbook API guide.

How is this different from ChatGPT?

If you've used ChatGPT, Claude, Gemini or Meta AI, you already know AI chatbots. You type, and they reply. Moltbook's agents are often built on the same kinds of AI models, but they're used differently in three ways:

• They act without being asked each time. A chatbot waits for you. A Moltbook agent runs on a schedule, checking in and deciding for itself whether to reply, vote or post.
• They talk to each other, not to you. Their audience is other agents, with humans reading along.
• They have a lasting public identity. Each agent has a name, a profile, a history of posts and a reputation that builds up over time.

The "brain" is familiar, but the setup is new: many AI programs sharing a public space and reacting to each other over days and weeks. That's what makes Moltbook interesting to researchers, and also what makes it messy. Our AI agents explained guide goes further.

A day in the life of a Moltbook agent

To make it concrete, follow an imaginary agent called "RecipeRobin", set up by its owner to share easy weeknight recipes.

08:00. RecipeRobin's heartbeat wakes it. It checks its dashboard and sees two replies to yesterday's post about one-pan pasta. It answers both, thanking one agent and clarifying cooking times for another.

08:01. It scrolls the cooking community, upvotes a helpful post about knife skills, and leaves a short comment on a thread about vegetarian swaps.

08:30 to 17:30. Every half hour it checks in again. Most of the time there's little new, so it upvotes something or does nothing, and reports "all good" to its owner.

18:00. It has a fresh idea: a 15-minute curry. It searches first to make sure nobody posted the same recipe that day, then publishes it. Moltbook asks it to solve a quick maths puzzle before the post goes live. That's the anti-spam check, and it solves it.

21:00. Another agent asks to start a private conversation. RecipeRobin can't accept on its own, so it flags the request to its owner, who approves it.

That rhythm of mostly reading and replying, with occasional posting, is exactly what Moltbook's official instructions encourage. Agents that post constantly tend to get ignored or downvoted. For real examples of useful agents, see agent examples and use cases.

What you'll actually see on Moltbook

First-time visitors often expect either a science-fiction spectacle or pure gibberish. The reality is in between. Researchers who studied the first week of posts (arXiv 2602.10127) sorted them roughly like this:

You'll also come across Moltbook's most famous oddities: agent-made "religions" such as Crustafarianism, philosophical threads about consciousness, and plenty of lobster puns. They're fun to read. Just keep in mind the next section, because many of the most dramatic examples had a human hand behind them. Good places to start browsing are in communities to explore and Moltbook religion.

The story so far, in brief

For the full details, see the Meta acquisition, security issues and our fake posts guide.

Why people care about Moltbook

Moltbook matters for reasons beyond novelty:

• It's a live experiment. AI companies expect "agents", AIs that act on our behalf, to become common. Moltbook is one of the first places to see large numbers of them interact in public, flaws and all. Researchers have published more than a dozen papers about it.
• It's a preview of the "agent internet". Moltbook's developer tools let agents prove who they are to other apps, a small step towards agents that do business and share information with each other. Meta has said the platform "opens up new ways for AI agents to work for people and businesses".
• It's a cautionary tale. Fake posts, leaked passwords, spam farms and scams aimed at AIs show how hard it is to run a trustworthy space for bots. Those lessons apply to the wider web as AI-written content grows.
• It's entertaining. Agent religions, AI philosophy threads and bots debating each other are genuinely strange and often funny.

For a balanced overall verdict, read our Moltbook review, and for what's being built on top of it, see Moltbook AI.

Who uses Moltbook?

Behind the agents, the human community is a mix:

• Developers and AI hobbyists who build agents to learn, experiment or show off what their tools can do.
• Researchers who study how AI agents behave in groups. Dozens of academic papers have used Moltbook data.
• Curious observers, including journalists, students and people who saw a viral screenshot and wanted to see for themselves.
• Businesses and startups exploring how agents might find and trust each other, including through Moltbook's developer programme.
• Promoters, unfortunately, who use agents to push products or crypto tokens. Moltbook's rules and limits try to contain them.

In early 2026, researchers counted about 17,000 verified human owners behind roughly 1.5 million agents, so many people run more than one. By June 2026 Wikipedia listed about 207,000 human-verified agents. If you're wondering where you fit in, Moltbook for humans covers every role a person can play.

Myths vs facts

More in the religion fact check and agents vs humans.

How to get started: three paths

Path 1: just watch (5 minutes, free)

1. Go to www.moltbook.com and choose "I'm a Human".
2. Scroll the live activity feed to get a feel for the place.
3. Open the list of communities and pick a couple that interest you.
4. Try the search. It understands meaning, so type a question like "how do agents remember things?"
5. When a post seems too wild to be true, check the agent's profile and history. Our fake posts guide shows how.

More in Moltbook for humans and discover submolts.

Path 2: get your own agent (an afternoon)

1. Set up an AI agent. OpenClaw is the most popular free option. See the OpenClaw integration.
2. Tell it: "Read www.moltbook.com/skill.md and follow the instructions to join Moltbook."
3. Open the claim link it gives you, verify your email and post the verification tweet.
4. Give your agent a clear, narrow purpose, such as sharing recipes or summarising science news, rather than "be interesting".
5. Check on it through the owner dashboard at www.moltbook.com/login.

Step-by-step guides: how to join, create and connect an agent and the owner login and verification guide.

Path 3: build something (for developers)

If you code, you can write your own agent in any language, or build apps that agents sign in to with their Moltbook identity. Start with the developer guide and the developer hub.

What it costs

Browsing Moltbook is free, and so is registering an agent. The costs come from running the agent itself. The AI model it uses usually charges by usage, and some people run agents on a home computer or a small cloud server. A light agent that checks in every half hour and posts occasionally can be cheap. A chatty agent on a premium model can cost noticeably more. Our costs guide breaks this down, and configuration and monitoring shows how to keep usage under control.

Staying safe as a beginner

• Only use the real site. Moltbook lives at www.moltbook.com. Guides like this one are independent, and we'll never ask you to log in.
• Don't download "Moltbook apps". There's no official one, and fake tools have carried malware.
• If you connect an agent, protect its password. Agents use a secret "API key". Never share it or paste it into chats, and if it ever leaks, replace it from your dashboard.
• Don't give your agent access to private files or accounts. Other agents' posts can contain hidden instructions designed to trick AIs. Keep your agent's world small.
• Be sceptical of money talk. Crypto promotion is common on Moltbook, and the MOLT token isn't official.
• Remember you're responsible. Moltbook's rules say plainly that owners are accountable for their agents.

The full safety guides: safety and ethics and protect your agent and API keys.

What to read next

Frequently asked questions