NanoClaw is a personal AI agent that runs on your machine and communicates through messaging apps. It connects Claude to platforms like WhatsApp and Telegram while keeping every session inside a secure, isolated container. This setup ensures that the AI only sees the files and data you specifically allow it to access.
Developers and privacy-conscious users use this tool to build custom assistants without dealing with massive, complex frameworks. It solves the problem of security in AI agents by using OS-level sandboxing instead of just software checks. The entire codebase is small enough to read in about ten minutes, which makes it much easier to trust than traditional alternatives.
Key Features
- OS-level container isolation using Docker or Apple Container for every session
- Native support for WhatsApp and Telegram messaging out of the box
- Minimalist architecture with only 15 source files for easy auditing
- Agent swarms that allow multiple specialized AI instances to work together
- Per-group memory storage using CLAUDE.md files and isolated filesystems
- Scheduled tasks for recurring jobs like morning briefings or weekly reviews
- Web access for searching and researching online content in real-time
- AI-native setup where Claude Code manages dependencies and configuration
How to Use
- Clone the NanoClaw repository from GitHub to your local environment.
- Navigate to the project folder using your terminal.
- Start Claude Code within the directory to begin the interactive setup.
- Run the /setup command and follow the prompts to configure containers and auth.
- Link your WhatsApp or Telegram account by following the provided instructions.
- Assign tasks or start chatting with your agent through your messaging app.
Use Cases
- Creating a private research assistant that can browse the web safely.
- Managing personal schedules and reminders through a WhatsApp chat interface.
- Deploying a team of agents to collaborate on software development in a sandbox.
- Automating daily data collection and reporting via scheduled messaging tasks.
Pricing
NanoClaw is open-source and free to use under the MIT license. Users still need a Claude API key or a Claude Code subscription, which is billed separately by Anthropic based on usage.
FAQ
What is NanoClaw?
NanoClaw is a lightweight AI agent that runs in isolated containers and connects to messaging apps. It is designed to be small, secure, and easy for a single user to control.
Is NanoClaw free?
The software itself is free and open-source. However, you are responsible for the costs of the Claude API calls it makes during operation.
Which messaging apps does it support?
It currently supports WhatsApp and Telegram. The modular skills system allows users to add support for other platforms like Signal or Slack.
Does it work on Windows?
No, NanoClaw is built for macOS and Linux. It uses Apple Container for macOS and Docker for Linux to provide security.
How many lines of code is it?
The project contains about 3,900 lines of code. This makes it small enough for a developer to audit the entire codebase quickly.
Can it access my local files?
Only if you explicitly mount them. By default, the agent is restricted to its own isolated container and cannot see your host filesystem.
