NanoClaw partners with Docker to secure AI agent development

On Friday, the open-source AI agent-building tool NanoClaw announced a partnership with Docker to integrate Docker Sandboxes, which are isolated environments for software execution. The deal comes about six weeks after NanoClaw was first introduced by its creator, Gavriel Cohen. The project is now transitioning from a personal tool into a commercial business under a newly formed company, NanoCo, where Gavriel Cohen’s brother, Lazer Cohen, serves as CEO. Docker, which counts nearly 80,000 enterprise customers, provides the container technology that underpins NanoClaw’s security model.

NanoClaw was developed as a direct, secure alternative to OpenClaw, a popular AI agent-building tool. OpenClaw has been widely reported to be a “security nightmare” due to its memory and account permission access, which makes it difficult to limit the tool’s access to data on a machine once installed. While OpenClaw is, by some estimates, sprawled across 800,000 lines of code, NanoClaw operates on just 500 lines of code. This minimal footprint allows developers to easily validate the codebase and run agents within isolated environments, preventing unauthorized access to sensitive local data.

The project has experienced rapid viral growth since its launch. About three weeks ago, an endorsement from AI researcher Andrej Karpathy helped propel the tool’s popularity. NanoClaw has now amassed 22,000 stars on GitHub, 4,600 forks, and over 50 contributors. Before focusing full-time on NanoCo, the Cohen brothers operated an AI marketing startup that they attributed was on track to hit $1 million in annual recurring revenue. They closed that startup about a week ago to build NanoCo. Reflecting on the project’s sudden scale, Gavriel Cohen said, “This is no longer my own personal agent that I’m running on my Mac Mini. This now has a community around it. There are thousands of people using it. Yeah, I said, I’m going to move over to the standard.”

Currently, NanoCo is operating on a friends-and-family fundraising round, according to Gavriel and Lazer Cohen. While the tool itself will remain free and open source, the founders plan to build a fully supported commercial product. Their strategy includes offering services such as forward-deployed engineers—specialists embedded directly with client companies—to help enterprises build and manage secure AI agents. This model aims to address security concerns for businesses deploying agents across communication tools like WhatsApp, which is to most of the world what Slack is to corporate America.

Why it matters

The integration with Docker signals a maturation of the AI agent ecosystem, as projects move away from the ‘wild west’ security models of early tools toward standardized, enterprise-grade infrastructure.