Don’t like ChatGPT or Gemini? A DIY-like AI assistant is here but there’s a catch
This is an AI chatbot that “lives” in your device and can be summoned in various messaging apps.
There are quite a few AI chatbots users can choose from, with their creators constantly updating them to make them more useful and intuitive.
Many of these chatbots run via applications, extensions, and integrations that let them be used with other platforms, increasing their use while we write emails, plan events, or look for directions in maps.
Recently, Peter Steinberger, the founder of PSPDFKit, now known as Nutrient, released a new chatbot called Clawdbot.
Clawdbot is an open-source AI assistant that drew attention for how customizable it is.
How does Clawdbot work?
Clawdbot isn’t like ChatGPT or Gemini – it doesn’t come in an app. Instead, it runs locally on a user’s device. The assistant can't be used on Mac, Linux, or Windows devices.
Users can try Clawdbot by downloading the source code the creator shared on GitHub.
While it might seem that all it takes for the program to work is to download and install it on one’s device, that isn’t quite the case, as users have to understand some technicalities for Clawdbot to work.
How is Clawdbot different from other AI assistants?
While other AI assistants have limitations, such as where they’re implemented and how much they can remember, Clawdbot was created to overcome them.
Besides being used on any device, it can also be used with any local models. Users can “talk to it” on various messaging apps, including WhatsApp, Telegram, Discord, Slack, and even iMessage.
The conversation on this topic is live. Join in the discussion.
When it comes to other AI agent features, Clawdbot can remember previous conversations and users’ preferences.
The assistant can also browse online, fill out forms, extract data from any webpage, run commands using locally stored data, read files, and more.
Is Clawdbot safe to use?
While the assistant has already been praised by some users, those considering installing it should also consider the privacy risks.
Once the AI agent is installed, it has full access to a user’s device, meaning it can read their files, control the browser, run scripts, and execute commands.
Even the installation instructions note that “running an AI agent with shell access on your machine is… spicy.”
“Clawdbot is both a product and an experiment: you’re wiring frontier-model behavior into real messaging surfaces and real tools. There is no ‘perfectly secure’ setup,” warns the security page of Clawdbot.
User reactions to Clawdbot: tricky to install, following privacy concerns
Besides already listed features of Clawdbot, one Reddit user noted that it's “the AI assistant that actually messages you first,” referring to its ability to send alerts and reminders when needed.
Their post about the AI assistant sparked an online debate, with some people still trying to figure out how to install it.
“Now I’m over even playing around with this thing for the rest of the day,” wrote one user after spending most of the day figuring out how to install the AI assistant.
“Is it that easy to set up one as a non-technical person?” asked another user.
Clawdbot: the AI assistant that actually messages you first
byu/jpcaparas inLocalLLM
The post also sparked debate about Clawdbot's safety, given that it has access to users' data.
“Clawdbot opens up the potential for a massive supply-chain attack that can steal or destroy everything on your machine and network, harvest and exfiltrate emails, crypto wallets, bank account and credit card details, SSN numbers, personal information, ssh keys,” wrote one user.
“From a security perspective, I have zero interest in this right now (LLM autonomy with access to my calendar, etc.),” shared another netizen who also noted that the benefits that Clawdbot provides are “minimal” as they can receive various reminders without the help of AI.
Security specialist has his doubts about Clawdbot
Jamieson O’Reilly, an offensive security specialist, shared his findings during the Clawdbot deployment.
The specialist explained that one of the main concerns is that “if not done right – people could be leaving their literal devices open waiting to be controlled by anyone on the internet.”
The problem arises when an AI agent is allowed to connect LLMs to various messaging apps and automation tools, potentially exposing these dashboards.
The specialist compared this to a case in which a person hires a butler, giving them a key to their home and access to their calendar, messages, and more.
While they do get the job done, one day the person finds a stranger from the street wandering in their home with their butler “cheerfully serving tea.”
undefined Jamieson O'Reilly (@theonejvo) January 25, 2026
Already in need of a rebrand?
If you’re looking to test out ClawdBot, it no longer goes by that name. It is now rebranded to Motlbot, with Clawd changing to Molty.
The change was prompted by Anthropic, perhaps due to name similarities between its Claude and ClawdBot.
🦞 BIG NEWS: We've molted!
undefined Mr. Lobster🦞 (@moltbot) January 27, 2026
Clawdbot → Moltbot
Clawd → Molty
Same lobster soul, new shell. Anthropic asked us to change our name (trademark stuff), and honestly? undefinedMoltundefined fits perfectly - it's what lobsters do to grow.
New handle: @moltbot
Same mission: AI that actually does…
According to the project’s developers, while the name changes, its mission stays the same, providing AI that “actually does things.”
Unlock more exclusive Cybernews content on YouTube.
