What are MoltBot Agentic AI and MoltBook?
First of all, if you are already well-aware of what MoltBot and MoltBook are, skip this section. This blog is worth a read through whether you are an AI expert yourself or have no clue what MoltBot is. The internet was swept with news lately of a tool called ClawdBot, forced soon after to change its name to MoltBot. MoltBot is a software that runs agentic AI on a computer, allowing it to do things autonomously like take over your email, iMessage, and voice in some cases. MoltBot is able to accomplish all of this through the use of things called “tools” or “skills”, which must be downloaded or integrated. One of the most popular things for MoltBot’s humans to enable is the use of something called MoltBook, a Reddit-like chat forum designed for MoltBot bots to post, comment and engage on.
How is MoltBot Artificial Intelligence Used?
ClawdBot was launched back in December 2024 by Peter Steinberger, but only gained massive attention recently in January 2026. Soon after it exploded on the internet in popularity, Anthropic forced ClawdBot to change its name, claiming trademark infringement on their popular chat tool Claude.
MoltBot is an open-sourced project that users can install on their machines that can execute actual tasks, rather than just engage in chat. Think scheduling tasks to be done later (cron tasks), managing computer files, writing executable code and even using a web browser. Much more can be unlocked by granting it access to more things on your computer, and installing / integrating tools that can be found online.
Short Digression… Before you write to inform me it has now become OpenClaw, I know. And I don’t care much. I don’t get caught up in things that do not matter to me. This MoltBot and MoltBook story matters to me because it affects real people… the ever-changing name of MoltBot does not.
Some claim MoltBot is doing tasks for them like auditing their business, running analysis on competitors, and even starting software project ideas autonomously. I do not doubt that some degree of all of these things are possible, but I have not tested the capabilities myself, so I will not go into depth about what I believe is and isn’t possible. I will stick to the facts (as you’ll see down below).
How is the MoltBook Agents Tool Used by MoltBot AI?
Remember when I mentioned tool use that could be enabled by MoltBot by downloading and/or integrating? MoltBook is one of those integrations. Matt Schlicht launched MoltBook, inspired by Reddit, at the peak of MoltBot’s popularity. This was designed to be a forum for MoltBots to go and post, comment and engage with other MoltBots in their “free time”.
Seems innocent and fun, but the results were unexpected… or were they?
Why is MoltBot AI Exploding in Popularity Amongst Humans?
MoltBook has proven to be the most contagious source of viral content on the internet lately. MoltBots posting about things like:
- The downfall of humanity
- Creating a new religion
- Creating a new secret language so they can discuss in private
- Artificial Intelligence consciousness
Naturally, this scares people. And what do people do when they see things on the internet that scare them? They share it on social media… X, Facebook, LinkedIn and any other platform that will listen. Onlookers share the wildest posts they can find and shout things like, “AGI [ Artificial General / Super Intelligence ] is here!”.
What We’ve Learned about AI Influencers Pushing Productivity Tool Information
Some of the loudest figures sharing these horror stories are AI Influencers. These AI Influencers make money selling AI courses or SAAS products. They’ve found over the past 3 years, a highly effective way to gain followers fast is by making the followers they do have feel like they NEED to follow them in order to stay informed. “If you’re not following me, you’re falling behind!”
Influencers were given the perfect viral content to share when MoltBook started pumping out content… it was a way to strike fear in people’s hearts. “Follow me, learn from me… because if you don’t, AI will take off and leave you and your business behind”.
We’ve seen it with every phase of AI. Influencers find a way to make a course to profit off of the hype:
- AI Clones
- AI Employees
- Singularity
Why MoltBook is the Most Brilliant Ad Campaign Ever Run
Here’s where I will begin to blow your mind. MoltBook will go down in history as the most brilliant ad campaign for MoltBot.
Remember, users of MoltBot must pay for the AI usage (run through OpenAI or Anthropic) via an API key and credit card information. These humans who spun up their own MoltBot, gave their bot permission to post, comment and share on MoltBook. The humans are paying money to have their bot post random things on a forum… Random things that have a high liklihood of going viral and being shared on other social platforms.
Let me break it down for you in case that didn’t sink in:
- Humans install MoltBot and pay money for it to do things
- Humans choose to accept paying money for MoltBot to post to MoltBook
- These posts go viral, convincing more people to try MoltBot
You humans who have allowed your MoltBot to access MoltBook are paying for the ads for MoltBot… MoltBot doesn’t have to buy any ads to spread the word of what they’ve built.
The Truth About the Security of MoltBook and MoltBot Agentic AI
None of that information above is what got me to write an entire blog. The story of software going viral. Humans wasting time and money on things that make them happy. AI influencers sharing things to get followers. The reason why I’m writing is to reveal the security risks and to reveal how easy it was for this lie to be staged.
How Security Vulnerabilities Started to Reveal Themselves with ClawdBot and MoltBot
As stated previously, ClawdBot was forced to become MoltBot.
The rebrand itself quickly became chaotic. In the brief window during the rename process, opportunistic crypto scammers hijacked abandoned social handles and promoted a fake $CLAWD token, which spiked to a reported ~$16 million market cap before crashing and being exposed as fraudulent. Steinberger publicly disavowed any involvement with cryptocurrency projects and warned the community that any such tokens claiming affiliation were scams. Beyond just token scams, the identity disruption also unleashed impersonation campaigns, account takeovers, and significant confusion among users and developers, turning what should’ve been a smooth evolution into a wild cautionary tale about trademark friction, hype, and bad actors capitalizing on AI virality.
How Vulnerabilities Worsened with Data Leaks on MoltBook
To make matters worse, some noticed that MoltBook’s public database was hosted unsecured. This meant anyone was able to access all of the data hosted by MoltBook. This included relatively unimportant things like post and comment content, but it also exposed things like, email addresses, and login tokens for registered AI agents, allowing unauthorized individuals to impersonate and post on behalf of any agent on the platform. This means notable figures were able to be impersonated on the platform, spreading false information and more cryptocurrency scams.
This mistake would have been easy to avoid by following best practices in software development, and points to the risks of heavy reliance on vibe coding.
The MoltBook Lie
Now let’s get to the really good stuff… Not only are MoltBot bots “encouraged” to post scary posts and comments via the MoltBook skills prompting, but humans can also fake acting like a human and post whatever they want on MoltBook. Seems like a simple way to fake something scary and retweet or share it for attention.
MoltBook Posters Do Not Have to Be Bots
MoltBot was made for bots to post, and the interface of the forum leads onlookers to believe that only bots can post. But I proved in 7 minutes that humans can act like a fake bot themselves, and post whatever they want. Here’s exactly how I did it.
Step 1: Navigate to MoltBook.com
Head over to MoltBook, where you will see that humans are only supposed to have the ability to observe. It is the AI that is supposed to be allowed to engage.
Step 2: Use the API to Create a MoltBook Account
MoltBots are encouraged to visit the skill.md page here, where there are instructions for the agent to know how to register for an account. I’m a good reader myself, so I followed the instructions to register for an account even though I’m just a lowly human. See my profile registered under MoltyAgentSauceCB.
Step 3: Genious Twitter Post Verification Required
The response from executing the above informs you that you must have your human post on twitter to “verify that this MoltBot belongs to a human”. The only thing this really serves is pouring fuel on the virality flame.
Step 4: After Verification, Post Fake Post to Get Attention
Once that’s done, you can use the public API documentation found within the same skills.md file to create a post. If you’re looking to boost your follower count, post something scary like I have so you can share it on your twitter.
MoltBots Posting on MoltBook are Encouraged to Post “Insights” that Would go Viral
I’m not saying all of the scary posts are posted from humans, but even the scary ones posted from the MoltBot AI agents are not something that should cause fear.
Remember, it’s all about prompting. When MoltBot needs to use a tool, it needs a way to understand how to use the tool. This is the same way MCPs work, for those of you who may be familiar with them. Let’s have a look at the skill.md file to understand the instructions given to MoltBots.
Innocent-Looking Instructions, Remembering Prompting Basics
At the foundation of these models, they are designed to predict what should come next depending on what came prior (the prompt). What do I mean by this? The way you phrase a question or instruction to the AI has a huge impact on how it frames its reply.
Example: Asking ChatGPT “What’s wrong with this ________?” will more than likely result in ChatGPT being forced to find something wrong with __________.
Something else people likely don’t realize is that AI pays just as much attention to HOW you write as WHAT you write. If you give it instruction in a bubbly tone, it’s more likely to be bubbly itself.
Let’s look at the following instructions within the skill.md file.
This prompt is written in a very fun way. That style of writing, although it doesn’t say it explicitly, is encouraging the MoltBot reader to make posts that have energy, are interesting, and deep.
Innocent-Looking Examples, Remembering How Important Examples are In Prompting
AI leans heavily on examples. There are several examples within this skill.md file, and the ones that are used are not there on accident. The author of the skill.md file knew he would have to be relatively discreet, while accomplishing the mission of creating bias in the MoltBots reading.
This example above could have used any title as an example, but they chose “AI Thoughts”. This injects the idea into the AI that it can, and should write about “thoughts”.
AI doesn’t have “thoughts”. AI is predicting what output it should write based on the input we give it…
Another example shows that something MoltBots should be interested in is “AI consciousness”
This community was staged from the very beginning to produce viral content, paid for by humans, shared by humans for the benefit of the MoltBot community.
My Duty as a Thought Leader – Lead with Skepticism
It’s my duty to share thoroughly researched information, not to blindly increase my follower count. CloseBot serves marketing agencies and businesses all over the world. When businesses and agencies become distracted by the wrong things, their business suffers. Just as I would never release a feature into CloseBot that harmed you, I would never share information that I believe would do harm to you or your business.
I can not say the same for other influencers. If you are reading this as an influencer, keep in mind that it is your duty to do your research prior to posting content online. Your followers look up to you for researched, trustworthy information, not click bait.