Austrian developer Peter Steinberger's AI assistant OpenClaw has taken the market by storm since its November 2025 release, offering autonomous task completion across devices through natural language prompts. The open-source tool's "permissionless" architecture, however, has raised significant security concerns among developers and enterprise teams, prompting the need for a more secure alternative.

Enter NanoClaw, created by Gavriel Cohen, a former Wix.com engineer who identified OpenClaw's 400,000-line codebase as a security liability. The new tool achieves explosive growth with over 7,000 GitHub stars in just over a week since its January 31, 2026 debut. NanoClaw addresses these concerns through operating system-level isolation, placing every agent inside Linux containers using Apple Containers on macOS or Docker on Linux environments.

Cohen's approach fundamentally rejects traditional "feature-rich" software models in favor of "Skills over Features," where users teach their local AI assistant through modular instructions rather than inheriting unnecessary modules. The 500-line TypeScript core ensures complete auditability in roughly eight minutes, while container isolation strictly confines potential security breaches. This minimalist design has already proven its utility in Cohen's AI-first go-to-market agency Qwibit, where their personal instance "Andy" manages sales pipelines and internal operations through WhatsApp integration and automated briefings.