NanoClaw¶
What it is¶
NanoClaw is a lightweight, AI-native personal assistant framework designed as a secure, containerized alternative to OpenClaw. It runs on the Claude Agent SDK and prioritizes codebase simplicity and OS-level isolation.
What problem it solves¶
It addresses the security risks and code complexity of heavy agent frameworks by providing a minimalist, container-first assistant that evolves through self-modification and composable skills.
Where it fits in the stack¶
Development & Ops / Personal Assistant. It is a lightweight agent framework for individuals and developers.
Typical use cases¶
- Secure, sandboxed AI assistance for personal local tasks
- Building custom multi-channel agents (WhatsApp, Telegram, etc.)
- Prototyping agent swarms in a low-complexity environment
Strengths¶
- Security-First: Native container isolation
- Minimalist: Small codebase, easy to understand and fork
- Self-Modifying: Can evolve its own features through code transformations
Limitations¶
- Not designed for complex enterprise multi-user scenarios
- Requires comfort with self-modifying code concepts
When to use it¶
- When you want a personal AI assistant that can be fully understood and customized (low code complexity).
- When you require strong security via Linux container isolation (Apple Container or Docker).
- If you prefer a "skills over features" model where the assistant evolves through code transformations rather than configuration.
When not to use it¶
- If you require a managed service or a complex, multi-user enterprise framework.
- If you are not comfortable with an assistant that modifies its own source code to add features.
Key Features¶
- Container Isolation: Agents run in sandboxed Linux containers, seeing only explicitly mounted directories.
- Skills-Based Architecture: Uses Anthropic Agent Skills to add capabilities (e.g.,
/add-whatsapp,/add-telegram) by dynamically rewriting the NanoClaw codebase. - Multi-Channel: Supports WhatsApp, Telegram, Discord, Slack, Gmail, and more.
- Agent Swarms: First personal assistant to support collaborative teams of agents.
- Small Footprint: Designed to be small enough for a single developer to understand the entire codebase.
Getting started¶
Requirements¶
- macOS or Linux
- Node.js 20+
- Claude Code
- Docker (or Apple Container on macOS)
Installation¶
git clone https://github.com/qwibitai/nanoclaw.git
cd nanoclaw
claude
/setup inside the Claude CLI.
Related tools / concepts¶
Sources / References¶
Contribution Metadata¶
- Last reviewed: 2026-04-18
- Confidence: high