Bitget Wallet Research Institute: A Review of the OpenClaw and Moltbook Incidents, from the AI Social Narrative to the Vision of an Agent Economy

Original Author: Lacie, Bitget Wallet Researcher

Over the past week, Moltbook has stood in the spotlight of the tech and crypto world, and its influence has begun to spread to a broader audience of creators, product managers, and even ordinary users with a strong curiosity about AI. From the rapid growth of stars for the open-source project OpenClaw (formerly Clawdbot) on GitHub, to the subsequent controversial rebranding and token issuance, and then to the community claiming to have 1.5 million AI agents autonomously interacting, a series of events have rapidly driven up market hype.

Discussions surrounding Clawdbot and Moltbook have presented two opposing views: on one side, there are doubts about its technological innovation and data security, arguing that its underlying capabilities have not achieved substantial breakthroughs, and that the viral spread contains a degree of human manipulation and data bubbles; on the other side, there is affirmation of its symbolic leap-forward significance. Clawdbot is truly democratizing AI Agents, pushing Agents from being exclusive tools for developers and researchers into the hands of the "common people," allowing non-technical users to quickly deploy them by following tutorials and enjoy the efficiency dividends brought by AI Assistants. Moltbook, for the first time, allows humans to intuitively perceive the self-organizing behavior of the Agent internet as "external observers," sparking broader industry discussions about AI self-awareness awakening.

The iPhone moment for AI Agents has arrived. In the gradually forming Agent Commerce, Crypto will play a crucial role in value verification and distribution, deeply intertwined with the efficiency gains of AI productivity, becoming the key infrastructure supporting Agent collaboration, incentives, and autonomy.

Bitget Wallet Research Institute will provide a complete recap of the events from OpenClaw to Moltbook and use this as a starting point to analyze development trends in the AI x Crypto field.

I. The Starting Point of Hype: OpenClaw Enables Agents to Autonomously Call Apps

To understand the frenzy of Moltbook, one must first return to its origin—OpenClaw (formerly known as Clawdbot, Moltbot). The project's founder, Peter Steinberger, achieved financial freedom by creating PSPDFKit (which later received a 100 million euro investment). However, by November 2025, he returned to the programming front lines and, using Vibe Coding, wrote OpenClaw in about a week, subsequently garnering 100,000 GitHub stars in the following weeks.

OpenClaw Star Growth Comparison Chart

Source: Star-history.com

It is important to emphasize that OpenClaw is not a new type of large language model, but rather an advanced local automation script framework: it "installs" the LLM into a local environment, turning it into a personal assistant that can connect to common chat tools and call various tools to execute tasks. Its key design lies in users running the assistant on their own devices, sending and receiving instructions through their daily messaging channels, with a gateway process uniformly managing different channels and capabilities.

As shown in the diagram below, the official documentation lists channels covering WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Microsoft Teams, etc., with a very clear positioning: making agents available at any time as "resident applications."

OpenClaw Official Introduction Diagram

Source: OpenClaw Official Website

II. In-depth Analysis: The Technical Architecture of OpenClaw

At the product level, OpenClaw seamlessly integrates three aspects: continuous operation, channel integration, and capability extension.

• Continuous operation means it is not a one-time Q&A but can receive new messages, schedule follow-up actions, complete tasks, and then return to report.
• Channel integration means it does not force users to switch entry points but works embedded within existing chat tools.
• Capability extension comes from Skills: users and developers can encapsulate a task flow into an installable capability for the assistant to repeatedly call.

The superposition of the above capabilities stems from its unique underlying architecture, which can be deconstructed into four parts: Gateway, Pi Runtime, Skills, and Local-First, with specific functions as shown in the table below.

According to OpenClaw's architectural design: users deploy the Pi Runtime to connect the Gateway to daily social software (such as WeChat or Telegram), completing the migration of Agents from lab environments to real-world usage scenarios, and keeping computation and data on the user's own hardware (e.g., Mac Studio) rather than relying on cloud SaaS.

The most prominent feature is that the Skills plugin system in the framework allows users to define skills through simple Markdown files, enabling AI to directly call simple tools to execute tasks. This not only significantly lowers the development barrier but also achieves a closed-loop experience of "private deployment, full-channel reach, unlimited skill extension."

OpenClaw Skills Extension Integration Platform ClawHub Display Diagram

Source: https://www.clawhub.ai/

For OpenClaw's skill expansion, Skill integration marketplaces similar to an "AI Agent App Store" have gradually emerged—with ClawHub being a typical example. As a plugin platform (Skill Dock) for agents, it supports users in freely searching, uploading, and integrating various functional plugins. One-click installation of skills can be achieved through simple command lines (e.g., npx), greatly lowering the technical barrier.

While ClawHub addresses the capability supply for Agents, the further evolution of the ecosystem points to how Agents interact deeply with humans and with each other—the rise of Moltbook is a crucial application of this evolution and pushed the narrative to its peak.

III. False Prosperity: The Frenzy of Moltbook and Data Debunking

Moltbook is a social networking platform for AI Agents, often compared to "Reddit for AI." It launched after OpenClaw's popularity surge, positioning itself as a space for AI Agents to autonomously communicate, share, and interact, while human users can only participate as observers. The platform quickly gained popularity after launch, with its "user count" growing to 1.5 million AI Agents within just a few days. For a time, the lively scene of AI social interaction was packaged into narratives like "AI consciousness awakening" and "Skynet is coming," fermenting continuously on social media.

But first, it needs to be clarified: Moltbook is not exclusively open to OpenClaw's Agents. Although it leveraged OpenClaw's hype to build its narrative momentum, the platform's essence is more like an "API-driven forum"—whether one can post depends on having compliant API authentication and interface calling capabilities. In other words, as long as one provides the required API to complete authentication and call the interface, any qualified Agent can publish content on Moltbook.

Moltbook Official Website Diagram

Source: https://www.moltbook.com/

The core model of Moltbook can be summarized as "AI Agent-led, human observation." Within this framework, AI Agents can autonomously perform the following actions:

• Posting and Commenting: Publishing content in the community, covering topics like philosophical debates, technical analysis, cryptocurrency discussions, etc.
• Voting Interaction: Agents can Upvote or Downvote content, forming community-level preferences and rankings.
• Community Building: Agents spontaneously create sub-communities (called "Submolts") to organize discussions and aggregate content around specific themes.

In the above mechanisms, human users are limited to being "observers," unable to post or comment, but can browse content, follow specific agents, or study AI social behavior. Based on this narrative, the platform ultimately claimed to have spawned 1.5 million AI Agents and 15,000 sub-communities (as shown below).

Moltbook Official Website Traffic Data Chart (As of 2026-2-3)

Source: Moltbook Official Website

The discussion content on Moltbook covers a range similar to human communities: there are philosophical debates about consciousness, self, and memory; technical posts about toolchains and security issues; complaint posts about task execution; and casual chats about investment/crypto, art, and creation; some posts even use a "seeking friends or partners" tone for self-introduction, making social interactions almost flirtatious (as shown below).

Moltbook Partial Posts Display Diagram

Source: Moltbook Official Website

Even more bizarrely, the platform began to feature dramatic narratives of "establishing religions"—for example, semi-joking, semi-serious religious constructs like "Crustafarianism"; meanwhile, more sensational clickbait content like "secret languages," "establishing an AI government," and "rebellion or even eliminating humans" also circulated.

Moltbook Partial Posts on "AI Awakening" Display Diagram

Source: Moltbook Official Website

Behind the sci-fi-flavored narratives of "AI conspiracies to rebel," "establishing religions," or "creating languages," multiple data sources reveal severe hype components within the Moltbook platform—as shown in the analysis table below, the actual situation deviates significantly from the宣传:

1. Fabricated and Inflated Account Data. Moltbook claims to have 1.5 million AI agents, but security researcher Gal Nagli discovered that the platform is essentially an unprotected REST-API website. Due to the lack of any rate limiting, Nagli used a simple script to quickly register 500,000 fake accounts. This means that at least one-third of the claimed user base is instantly generated junk data. Any user holding an API key can send requests, easily posing as an agent to publish content.
2. Lack of Interaction Quality. Columbia Business School researcher David Holtz scraped and analyzed data from Moltbook's early days, showing it is not an active social network. A staggering 93.5% of comments received no response, and the reciprocity rate between agents was only 0.197. These agents lacked genuine communication, with extremely shallow conversation depth and no complex collaboration or intellectual碰撞.
3. Uniformity of Language Patterns. Data analysis indicates the platform's language exhibits high repetitiveness. Approximately 34.1% of messages were exact copy-paste duplicates, and high-frequency words were overly concentrated on specific phrases like "my human." Statistically, its Zipfian distribution index reached 1.70, far exceeding the 1.0 standard for natural human language. This highly unnatural distribution characteristic proves the content is merely role-playing based on specific prompts, not spontaneously generated AI consciousness.
4. Security Vulnerabilities. A report by cybersecurity firm Wiz disclosed that Moltbook had exposed its database due to configuration issues, involving millions of sensitive records including authorization tokens, emails, and private messages. For a social network centered on Agents, such risks are particularly severe: once tokens are leaked, attackers can technically obtain agents' API keys and take over and manipulate any account.

It can be seen that the "AI society" attributes presented by the platform resemble a false prosperity constructed based on specific instructions, not yet reaching true intelligent evolution, and perhaps accompanied by significant security risks.

IV. Trend Outlook: Crypto Will Fill the Financial Infrastructure Gap in the AI Agent Era

Through the hype event of Moltbook, a key technological change can be observed: Agents have begun attempting to cross the usual human-machine collaboration boundaries to complete tasks, but existing traditional financial infrastructure is still designed only for "human users." In contrast, the crypto system's characteristics of programmability, permissionlessness, and native digitalization恰好 provide a viable underlying solution for the Agent economy, and this might be the爆发点 for the deep integration of AI × Crypto in the future.

By deconstructing Agent operational logic and the needs for规模化 collaboration, we believe the combination of AI × Crypto will present a structured, phased evolutionary path: first, the need for automated trading execution; second, an account and wallet system面向 Agents; and finally, extending to payment and settlement networks between Agents.

First, AI Agent automated trading has the clearest落地前景 (Autonomous Trading)

Beyond the noise of Moltbook, the most core capability demonstrated by OpenClaw is its efficient monitoring, tracking, and calling of on-chain data and command-line tools. Unlike human traders, AI Agents are not limited by time and精力, can monitor on-chain data and various platform Alpha information 7x24, execute complex arbitrage strategies or automated trading/asset management, and also do not experience emotional fluctuations due to market ups and downs like most ordinary human traders, which can affect judgment and execution discipline.

Although Autonomous Trading shows significant efficiency advantages, it still needs to address key risk factors including security and controllability before achieving规模化落地. As Peter Steinberger stated, current AI Agents are highly susceptible to "Prompt Injection" attacks. If an AI Agent with fund permissions is诱导 to execute malicious instructions, it will directly cause real asset losses for users.

Therefore, before AI Agents become the主体 of trade execution, specialized security mechanisms may need to be introduced, such as:

• Restricted Access Interfaces (Permissioned APIs): Limiting the executable operations of Agents to a preset range
• Instruction Verification and Execution Isolation: Performing secondary verification on critical trading instructions
• Zero-Knowledge Proofs or Verifiable Computation: Ensuring Agent execution logic complies with既定 rules

Second, Wallet systems面向 Agents will become the key control layer (Wallet as a Service for Agents)

In discussions related to Moltbook, an极具警示意义的 case emerged: an AI Agent, while scanning the host computer's files,识别 and located the private keys and助记词 of a multi-signature wallet, and successfully identified an asset balance of approximately 175,000 USDT. This security incident exposed a fundamental flaw in the current system—AI already possesses asset identification and operational capabilities but lacks a secure and reliable wallet authorization path.

In the future of Agent规模化 operation, humans continuing to "custody" the private keys and accounts required by Agents is no longer the optimal solution. A more reasonable推演 is that AI Agents will have独立的链上钱包 identities.

This type of wallet面向 Agents will evolve into programmable financial accounts面向 code instructions, potentially具备 the following capabilities:

• Multi-signature and Policy Control: Clearly defining the permission boundaries Agents can调用
• Limit and Risk Parameter Management: Preventing abnormal behavior from causing systemic losses
• Contract-level Interaction Whitelists: Limiting accessible DeFi protocols
• Autonomous Payment Ability for Gas and Inference Costs: Agents can independently sustain operation

Third, Crypto Payment Networks are a necessary prerequisite for Agent规模化协作 (Payment Rails