Gavriel Cohen’s encounter with OpenClaw, initially known as Clawd Bot, marked a pivotal moment that ultimately led to the creation of NanoClaw and NanoCo AI. What began as a developer’s attempt to rekindle his coding passion through Anthropic’s Claude Code quickly evolved into a deep-seated concern about the security, maintainability, and scalability of the open-source project. Cohen, a former developer at the prominent no-code website development platform Wix, recognized a critical void in his workflow, a void that OpenClaw, despite its initial promise, failed to fill. His subsequent development of NanoClaw is a testament to his commitment to building robust, secure, and user-friendly autonomous agent infrastructure, addressing the fundamental flaws he identified in its predecessor.
Cohen’s journey into the world of AI-driven automation began in March 2025. While taking a break from active development to focus on a marketing project, he found himself drawn back to coding by Anthropic’s Claude Code. This exploration, intended to sharpen his coding instincts, led him to discover Clawd Bot, the nascent form of what would become OpenClaw. "I had my first little go," Cohen recounted to The New Stack. "I installed it, connected it to my WhatsApp, sent a few messages back and forth. So I came to it really from a need as we were building an AI native marketing agency." This initial engagement, driven by a practical need for automation within his marketing endeavors, laid the groundwork for his subsequent critical evaluation of the platform.
The realization that OpenClaw was not the solution he envisioned dawned on him almost immediately. During the setup process, which offered a selection of packages to install, Cohen was taken aback to find his own independently developed package, NanoPDF, listed among the options. "Why did they include that tiny package?" he mused, expecting to see only well-established, highly reviewed third-party tools. His proprietary NanoPDF package, he knew, had limited user adoption and had not been updated in months, raising immediate questions about the curation and quality control within OpenClaw. This discovery, while perhaps initially surprising, was merely the first of several red flags.
The issues compounded as Cohen delved deeper. Within the first couple of days of using OpenClaw, a scheduled job failed to execute. Examining the logs to diagnose the problem, he uncovered a significant privacy breach: the logs contained messages from all connected WhatsApp groups, not just the specific one he had intended to integrate. This revealed a profound lack of segmentation and a concerning disregard for user privacy, issues that are particularly alarming in the context of AI agents handling sensitive communications.
The Unwieldy Codebase: A Fatal Flaw
Cohen’s concerns escalated dramatically upon examining the core architecture of OpenClaw. The project had ballooned to an unmanageable size, comprising approximately half a million lines of code. This immense codebase, according to Cohen, represented a fundamental flaw, rendering the project unwieldy and difficult to maintain, even for an open-source initiative with numerous contributors. By February of the same year, OpenClaw had accumulated over 3,000 unresolved pull requests, a clear indicator of its escalating maintenance burden.
"But most importantly, I looked at the code base, and it’s like a half a million lines of code," Cohen emphasized. For his marketing business, which operated with a lean team of three employees, the inability to securely connect OpenClaw to customer data and build a reliable business infrastructure on top of it was a deal-breaker. The complexity and apparent lack of stringent development standards made it impossible to integrate with sensitive customer information or scale the operations effectively.
The realization that OpenClaw was not a viable platform for his business needs prompted Cohen to take decisive action. "So Cohen did the one thing he knew he could do. ‘I sat down to build NanoClaw. I had to make this super small because in order for anybody who cares about security to use it, they’re going to have to be able to look over the code and actually see what’s going on and be OK with it.’" This commitment to a minimalist, transparent, and secure design became the guiding principle for NanoClaw.
Architecturally, Cohen identified four fundamental capabilities essential for an AI agent: a coding agent capable of writing code and executing Bash commands, a persistent environment session, integration with a messaging application, and internet connectivity. He reasoned that all advanced functionalities could be built upon these core pillars, enabling proactive operation through scheduled jobs rather than mere reactive responses. In contrast to OpenClaw’s sprawling codebase, Cohen aimed to develop a "claw agent" that could be as lean as 25 lines of code, demonstrating a stark difference in design philosophy and efficiency.
The half-million-line codebase was not just a technical challenge; it was a strategic impediment. Cohen’s assessment was blunt: "I think it was fundamentally flawed from the beginning, and the fatal flaw is half a million lines of code." A subsequent check revealed that OpenClaw’s codebase had grown to over 800,000 lines. Even after OpenAI’s acquisition of OpenClaw, the usability issues persisted, and the community engagement dwindled, underscoring the deep-seated problems within the project’s architecture and development trajectory.
Building from Scratch: The Genesis of NanoClaw
Cohen’s approach to building NanoClaw was one of deliberate reinvention, starting from a clean slate. Instead of analyzing OpenClaw’s architecture for inspiration or correction, he focused on defining the core requirements from the ground up. "When I actually sat down to write NanoClaw, I just started from an empty project, and I didn’t tell Claude Code to go look at OpenClaw," Cohen explained. "I just described the capabilities. I said I want a messaging app; a coding agent; I want it in a loop and I want memory and I built it from scratch."
This foundational approach allowed Cohen to prioritize security and efficiency from the outset. A critical decision was the implementation of isolated containers. "I can’t just have it running on my machine with an autonomous agent able to do everything, so I put it in an isolated container," he stated. This architectural choice not only enhanced security by preventing unauthorized access and mitigating potential damage from compromised agents but also provided a clear market differentiation.
Interestingly, NanoClaw’s initial containerization efforts leveraged Apple’s nascent native container capabilities. Cohen, working on a Mac mini, was aware of this feature and found it effective for initial development. However, as NanoClaw gained traction, the community expressed a preference for the industry standard, Docker. "Everybody wondered, ‘What are you doing with Apple containers? Docker is the default; it’s a standard that works everywhere.’ So I supported both and then within a few weeks I just changed the default to Docker containers," Cohen recalled. This adaptability underscores NanoClaw’s responsiveness to user needs and industry best practices.
To further bolster security and streamline credential management, NanoClaw partnered with OneCLI. This collaboration addressed a significant challenge: securely managing sensitive API keys and tokens required for agents to interact with external services. "Before we had that partnership, there was a major issue with credentials entering the agent environment that we were trying to prevent – but, for example, you need to get the Anthropic token in place so that the agent can connect," Cohen noted. The partnership with OneCLI provided a robust credential and proxying layer, offering a more sophisticated solution than NanoClaw’s initial in-house proxying system, which had become complex to manage. This integration, along with human-in-the-loop approval mechanisms and policy enforcement for agent actions, significantly enhanced the platform’s enterprise readiness.
Cohen also highlighted the strategic integration of Vercel’s Chat SDK. This open-source package provides standardized connections to a wide array of messaging applications, aligning with NanoClaw’s philosophy of avoiding redundant development. "One of the core philosophies I came to NanoClaw with is don’t reinvent the wheel. Why should everybody be re-implementing the same integrations with messaging apps – let’s have one library that we all congregate around," he asserted. This collaborative approach fosters a more unified ecosystem for AI agent development.
Redefining the Landscape: NanoClaw’s Enterprise Ambitions
In the competitive landscape of autonomous AI agents, Cohen believes NanoClaw is well-positioned to inherit and redefine the mantle left by OpenClaw. The company’s focus on addressing "the big hairy problems and challenges that stand between using autonomous agents with their full power, while doing it in a way that’s safe," including credential management, human oversight, and agent isolation, sets it apart.
Looking ahead, NanoClaw is strategically targeting large enterprise businesses with stringent security, compliance, and regulatory requirements. "In the next few months, we are looking at large business enterprises that have strict security requirements, compliance requirements, regulatory requirements, etc.," Cohen stated. The platform’s current capabilities, allowing secure connections to email and calendar services with mandatory approvals for sensitive actions, are key enablers for this market entry.
The journey from identifying the flaws in a sprawling, insecure open-source project to building a lean, secure, and scalable platform like NanoClaw is a compelling narrative in the rapidly evolving field of AI. While competitors vie for dominance in the wake of OpenClaw’s challenges, NanoClaw’s deliberate design choices, focus on security, and strategic partnerships position it as a strong contender to lead the next generation of autonomous AI agents. The company’s commitment to providing powerful yet safe autonomous agent solutions appears to be the key to its future success in the enterprise sector and beyond.
