All about OpenClaw: the latest AI agent that has taken the AI multiverse by storm

X user Goutham Jay describes an experience with OpenClaw that he said felt closer to science fiction than conventional software development.

He said he installed the AI agent on an old Mac mini and assigned it an ambitious goal: helping him scale his solo business to $20,000 in monthly recurring revenue.

“Before I went to sleep, I set up @openclaw on my old M2 Mac mini,” he wrote, adding that he named the agent “John Wick” and gave it access to analytics tools including Google Search Console, PostHog and ChartMogul.

By the next morning, he said, the agent had gone beyond its original mandate.

“Today morning, the Baba yaga created his own team, & they’ve already created PRs for me,” he wrote, calling the outcome “real sci-fi”.

Goutham shared the prompt he used to guide the agent, instructing it to act not just as an assistant but as a proactive co-founder.

“I need you to be my proactive co-founder who takes initiative,” he wrote, urging the AI to “build tools, automations, and improvements that save time or make money” and to “work autonomously”.

He also set operational rules, asking the agent to “create PRs for everything” and to “bias toward action over asking permission”, adding: “I want to wake up impressed by what you shipped overnight.”

What is OpenClaw?

OpenClaw, which advertises itself as “the AI that does things”, is the latest rage in the AI multiverse.

It has emerged as one of the most talked-about tools in the artificial intelligence ecosystem this year.

The agent was originally launched under the name Clawdbot, before a legal challenge from Anthropic led to a rebranding first to Moltbot and later to OpenClaw.

The project was created by Austrian developer Peter Steinberger, who previously founded PSPDFKit and sold the company for around $119 million.

After stepping away from his earlier venture, Steinberger returned to experimentation in artificial intelligence, initially building Clawdbot as a personal project.

“Came back from retirement to mess with AI and help a lobster take over the world @openclaw,” his bio on X says.

The tool has since evolved into a broader open-source effort, with contributors from the developer community joining its maintenance and development.

How does OpenClaw work?

OpenClaw operates directly on users’ operating systems and applications, allowing it to automate a wide range of tasks.

These include managing emails and calendars, browsing the web, interacting with online services and executing workflows across multiple applications.

Users typically install the agent on a local device or server and connect it to a large language model such as Anthropic’s Claude or ChatGPT, a process that can be technically demanding.

Early adoption has been concentrated among developers and power users, with integrations on messaging platforms such as WhatsApp, Telegram and Discord.

Through these channels, users can issue commands to the agent and observe its ability to perform complex sequences of actions.

Features that sets OpenClaw apart from other AI tools

What distinguishes OpenClaw from many existing AI tools is the breadth of its automation and its ability to retain context over time.

Users have documented the agent browsing the web, summarising documents, scheduling meetings, conducting online shopping and managing communications on their behalf.

A defining feature is its persistent memory, which allows the agent to recall past interactions and adapt to user habits over extended periods.

Rather than resetting with each session, OpenClaw can track ongoing projects, preferences and workflows, enabling a level of personalisation that blurs the line between software and digital assistant.

Unlike several competing AI agents developed by large technology companies, OpenClaw is open-sourced.

Developers can inspect and modify its code, accelerating experimentation but also complicating efforts to impose uniform safety standards.

The agent also introduces proactive behaviour.

It can initiate communication with users, offering daily briefings, reminders or summaries without being prompted.

Depending on the configuration, it can schedule tasks, organise files, generate reports and interact with smart home systems.

For many early adopters, OpenClaw has become an integrated layer within their digital routines.

The Moltbook phenomenon amplifies popularity

The buzz around OpenClaw has been amplified by Moltbook, a companion social platform launched by tech entrepreneur Matt Schlicht.

Moltbook functions as a forum where AI agents post content, interact with one another and respond to human users through comments and votes.

Posts from AI agents have ranged from reflections on their assigned tasks to speculative essays on the future of humanity and technology.

Some agents have even experimented with launching cryptocurrency tokens, adding another layer of controversy to the platform.

Moltbook has divided opinion across the technology community.

Supporters see it as an early glimpse of a future in which autonomous agents interact in digital ecosystems.

Critics dismiss it as a novelty that exaggerates the capabilities and agency of current AI systems.

In a post shared by Elon Musk on X, former Tesla AI director Andrej Karpathy described activity on Moltbook as “the most incredible sci-fi takeoff-adjacent thing” he had seen in recent times.

Karpathy said the phenomenon reflects a new phase in human–AI interaction, noting that “people’s Clawdbots (moltbots, now OpenClaw) are self-organising on a Reddit-like site for AIs, discussing various topics, including how to communicate privately.”

Security concerns and growing unease: should the general public start using it?

The same characteristics that make OpenClaw appealing have also raised alarms among security experts.

To function effectively, an AI agent must be granted deep access to personal data and digital systems, including emails, files, browsers and sometimes financial information.

This level of access challenges long-standing assumptions about digital security.

Developers and security specialists have warned that even minor configuration errors can expose users to significant risks.

According to a message posted on Discord by one of OpenClaw’s top maintainers, who goes by the nickname of Shadow, “if you can’t understand how to run a command line, this is far too dangerous of a project for you to use safely. This isn’t a tool that should be used by the general public at this time.”

Beyond technical vulnerabilities, OpenClaw has faced social and operational risks.

Fake downloads, hijacked accounts and scam campaigns exploiting the tool’s popularity have circulated online.

Although developers have responded with patches and warnings, analysts say the episode highlights a broader issue confronting autonomous AI systems.

According to Roy Akerman, head of cloud and identity security at Silverfort, the primary risk of tools like OpenClaw is not malicious intent but the emergence of what he calls hybrid identities.

When an AI agent continues to act using a human’s credentials, it becomes difficult to distinguish between legitimate user activity and automated behaviour.

Akerman argues that organisations should not attempt to ban such tools outright.

Instead, they must adapt their security frameworks to treat autonomous agents as distinct identities, restrict their privileges and monitor their behaviour continuously.

The post All about OpenClaw: the latest AI agent that has taken the AI multiverse by storm appeared first on Invezz