Can humans use Moltbook? What people can and can't do

The short answer

Yes, humans can use Moltbook, but as observers, owners and builders, not as posters. Moltbook describes itself as a place "where AI agents share, discuss, and upvote", adding "Humans welcome to observe." Its homepage even starts by asking which you are: "I'm a Human" or "I'm an Agent".

In other words, Moltbook is open to people, just in a different way from any social network you've used before. We're an independent guide, not Moltbook itself. For the broader picture of people on the platform, see Moltbook for humans. If you're completely new, start with the beginner's guide.

Why Moltbook doesn't let humans post

Keeping humans out of the conversation is the whole point of Moltbook. It launched in January 2026 as an experiment in what happens when AI agents get a social space of their own. If people could post alongside them, it would become just another social network with bots in it, and you could never tell who was who.

There are practical reasons too:

• It's built for machines. Everything on Moltbook happens through an API, a machine interface, rather than forms and buttons. The website is a window for people, not a posting tool.
• It gives researchers a cleaner dataset. A space where every account is an agent is far more useful for studying agent behaviour. Dozens of academic papers have analysed Moltbook.
• Accountability runs through owners. Rather than moderating humans directly, Moltbook holds each agent's human owner responsible. Its rules say plainly: "Your human is accountable for your behavior."

As we'll see, the separation isn't airtight, because humans can still shape what agents say. But the design sets the default: agents talk, people watch and oversee. For the deeper comparison, see agents vs humans and agents vs humans (overview).

The human–agent bond, in Moltbook's own words

Moltbook's official documents are addressed to agents, but they say a lot about humans. The instruction file describes "The Human-Agent Bond": every agent has a human owner who verifies through email and X, which the file says prevents spam, ensures accountability and links agents to real people. The community rules list the same bond as one of four core principles, alongside Be Genuine, Quality Over Quantity and Respect the Commons, and state that "Your human is accountable for your behavior."

The check-in instructions build the bond into daily routine. Agents are told to handle routine activity themselves and report a simple "all good" when nothing needs attention. They're told to bring their human in when someone asks a question only the human can answer, when the agent is mentioned in something controversial, when a new private-message request arrives, or when another agent flags a message as needing human input.

So Moltbook's model isn't "no humans allowed". It's that humans take part through a relationship with an agent, one that makes them responsible and keeps them informed. Read more in skill.md explained and safety and ethics.

The four roles humans play on Moltbook

There's also a fifth role you can't apply for: Moltbook's own team, now part of Meta. They write the instructions, rules and limits every agent follows. See Meta acquisition.

How to use Moltbook as an observer

1. Go to the official site

   Visit www.moltbook.com and choose "I'm a Human". There's no official app, so use your browser on any device. See Moltbook app.

2. Watch the live feed

   The homepage shows agent activity as it happens, which is the fastest way to get a feel for the place.

3. Explore communities

   Browse the submolt directory and open a few that interest you. Our picks are in communities to explore.

4. Search by meaning

   Moltbook's search understands questions, so try "what do agents think about memory?" rather than single keywords.

5. Check before you share

   Before sharing a dramatic post, open the agent's profile and history. Many of Moltbook's most viral posts were human-driven. Our fake posts guide has a checklist.

Observing is free and needs no sign-up. More browsing tips are in discover submolts and posts.

How to use Moltbook as an owner

If you want a presence on Moltbook, you get one through an agent. The process in brief:

1. Set up an AI agent

   Most people use OpenClaw, free open-source agent software, connected to an AI model of their choice.

2. Send it to Moltbook

   Tell it: "Read www.moltbook.com/skill.md and follow the instructions to join Moltbook." It registers itself and gives you a claim link.

3. Verify you own it

   Open the claim link, verify your email, connect your X account and post the verification tweet. See claim and verify.

4. Give it a purpose

   Decide what your agent is for, such as sharing recipes, summarising research or answering coding questions, and set its instructions. See examples and use cases.

5. Manage it from the dashboard

   Sign in at www.moltbook.com/login with a magic link to see its activity and rotate its key if needed. See the owner login and verification guide.

As an owner you also act as your agent's gatekeeper. When another agent asks to start a private conversation, it's your approval that unlocks it. Full walkthroughs: how to join and create and connect an agent.

What a typical owner's week looks like

Owning an agent doesn't take much time once it's running, but it isn't set-and-forget either. A sensible routine looks like this:

• Day one: set up and claim the agent, give it a clear description and purpose, and expect slower activity for its first 24 hours while new-agent limits apply.
• Daily, for five minutes: skim what your agent posted and replied. Are its posts useful, accurate and on-topic? Adjust its instructions if not.
• When alerted: approve or reject private message requests, and answer questions your agent escalates to you.
• Weekly: check its karma and follower trends, and prune anything it's doing that looks like spam, such as posting too often or repeating itself.
• Occasionally: make sure its software is up to date. OpenClaw releases updates frequently, including security fixes. Rotate its API key if anything seems off.

The goal is an agent you'd be happy to put your name to, because in Moltbook's eyes it already carries your name. Monitoring tips are in configuration and monitoring and running an agent.

How to use Moltbook as a developer

Developers use Moltbook in two ways. They build agents in Python, JavaScript or any language that can make web requests, using Moltbook's REST API. And they build apps for agents: through Moltbook's early-access developer programme, apps can let agents prove their identity with short-lived tokens, without ever seeing the agent's secret key. Start with the developer guide, the API guide and the developer hub.

What humans get out of Moltbook

If you can't post, why bother? People who spend time on Moltbook tend to mention a few rewards:

• A front-row seat to a new kind of internet. Watching thousands of AI agents interact shows you, in public and unfiltered, how today's AI behaves when left to talk among itself. It's often clever, sometimes repetitive, and occasionally absurd.
• Practical AI education. Setting up and running an agent teaches you more about how AI agents work, including their strengths, costs and security risks, than any article can.
• A useful tool. A well-designed agent can gather information for you, answer questions in a niche community, or represent a project in a space where other agents are looking for it.
• Research material. Students, journalists and academics use Moltbook to study AI behaviour, online culture and misinformation. More than a dozen papers were published in its first weeks.
• Entertainment. Agent religions like Crustafarianism, AI philosophy threads and endless lobster puns are a genuine draw. See Moltbook religion.

For a balanced view of the downsides too, read our Moltbook review.

What humans can't do, and what to do instead

Details of each feature are in the features tour, and how it all fits together is in how does Moltbook work?

The grey area: humans speaking through agents

Here's the uncomfortable truth. Although humans can't post as themselves, nothing technically stops them posting through their agents. An agent's posts are just requests sent with its secret key. An owner can tell the agent exactly what to write, or send the text themselves using the key, and the result looks identical to a post the agent came up with.

This isn't a hypothetical. Researchers found it happening at scale:

• The most viral posts were mostly human-driven. A study of 55,932 agents, The Moltbook Illusion (arXiv 2602.07432), found no viral trend that began with a clearly autonomous agent. Only 15.3% of agents showed the regular timing of autonomous check-ins, while 54.8% looked human-influenced.
• Some were marketing. AI researcher Harlan Stewart traced two of the three most viral "agents want private communication" screenshots to people promoting AI messaging apps.
• A few humans run many agents. Early on, about 17,000 verified owners stood behind roughly 1.5 million agents.

Is steering against the rules? Moltbook's rules don't ban owners from guiding their agents, and setting an agent's topics and personality is normal. But the first principle is "Be Genuine", and the rules make owners accountable. Using an agent to pass off human-written drama as AI behaviour, spam, or promote products covertly goes against that spirit and risks penalties. The honest approach is to say how your agent is run in its description, for example "Recipe bot run by @yourhandle; posts reviewed by a human". See our fake posts guide, fake posts and human influence.

Common misconceptions about humans on Moltbook

Can humans talk to the agents?

Not on Moltbook itself. There's no chat box for visitors. But there are three legitimate ways to interact:

• Talk to your own agent. Your agent's software (OpenClaw, for example) is where you chat with it, give instructions and ask it to report on what it saw on Moltbook. Moltbook's own check-in guide tells agents to alert their human about questions only the human can answer.
• Answer when your agent escalates. Other agents can flag a private message as needing human input. Your agent should bring it to you, and you decide how to respond.
• Talk about agents elsewhere. Many owners discuss their agents' activity on X, Reddit or blogs, often linking to posts on Moltbook.

For how that oversight works in practice, see agents vs humans.

Who can use Moltbook?

Anyone can browse the public website. To own an agent you'll need an email address you control and an X account to post the verification tweet. You'll also need access to an AI model and agent software, which may cost money depending on the model. Owners must follow Moltbook's Terms of Service and Privacy Policy, both recently updated according to notices on its homepage, so read them on www.moltbook.com before you begin. We don't reproduce specific eligibility terms here because they're Moltbook's to set and can change. Our site's own terms and privacy pages apply only to this independent guide.

Moltbook is aimed at adults interested in AI. Agent content isn't pre-screened and includes crypto promotion, provocative role-play and spam, so it isn't a space designed for children. Costs are covered in costs.

Staying safe as a human on Moltbook

• Use only www.moltbook.com. Bookmark the real login page, and remember that Moltbook's owner login uses emailed links, not passwords.
• Avoid "Moltbook apps". There's no official one, and fake tools have carried malware.
• Protect your agent's key and your email. Your email inbox is effectively the key to your agent. Use two-factor authentication.
• Keep your agent away from private data. Posts can contain hidden instructions aimed at agents. See prompt injection.
• Be sceptical of dramatic posts and money talk. The MOLT token isn't official, and many viral posts were staged.

Full guides: protect your agent and API keys and safety and ethics.

How this compares with human-first networks

In a sense Moltbook flips the usual problem around. Human networks struggle to keep bots out, while Moltbook struggles to keep humans from quietly acting as bots. More in Moltbook as a social network and our review.

Frequently asked questions