NanoClaw: A 500‑Line Minimal Claude Bot Running in Apple Containers

Problem with OpenClaw

OpenClaw’s repository contains more than 52 modules, eight configuration files, and over 45 dependencies. The resulting code‑base size causes slow startup—tens of seconds on an M2 Mini, comparable to launching Photoshop on an old desktop.

NanoClaw’s Minimalist Refactor

NanoClaw reduces the core functionality to roughly 500 lines of code, making the entire codebase readable within minutes. The author describes it as a “personal Claude assistant running in an Apple container.”

Container‑Based Security Design

Execution occurs inside an Apple‑provided container, limiting the assistant’s access to the host system and thereby reducing potential attack surfaces. The design sparked discussion: some users argue that an AI assistant’s value lies in unrestricted sandbox access, while others warn that excessive openness can introduce security vulnerabilities.

Docker Support for Linux

Recent commits add Docker support, enabling Linux users to run NanoClaw. The commit history shows ongoing security hardening, including limits on container output size and a whitelist for safe mount points.

Implications

NanoClaw demonstrates a concrete approach to mitigating code bloat in AI assistants, but the author notes that “nano” projects often experience rapid rise and fall. Developers can use NanoClaw as a learning reference while remaining cautious about heavy reliance.

Project repository: https://github.com/gavrielc/nanoclaw