NanoClaw is a lightweight, containerized AI assistant designed for secure and customizable interaction with various messaging platforms and AI models. Its primary purpose is to provide users with a secure and easily understandable platform for managing AI agents that can interact with services like WhatsApp, Telegram, Discord, Gmail, and more. Unlike more complex alternatives like OpenClaw, NanoClaw prioritizes simplicity and security by running agents within isolated Docker containers. This approach minimizes the attack surface and allows for granular control over agent access and permissions.
The core functionality of NanoClaw revolves around enabling users to create and manage AI agents that can perform various tasks, such as responding to messages, scheduling tasks, and accessing web content. The system supports multiple messaging channels, allowing users to connect their agents to platforms like WhatsApp, Telegram, Discord, and others. These channels are added on demand through "skills," which are essentially modular components that users can install to extend NanoClaw's capabilities. This modular design keeps the core system lean and allows users to customize their setup to their specific needs.
A key feature of NanoClaw is its focus on security through containerization. Each agent runs within its own isolated Docker container, preventing unauthorized access to the host system and other agents. This isolation is further enhanced by the use of OneCLI's Agent Vault, which securely manages API keys and credentials, ensuring that agents never hold raw authentication information. This architecture provides a robust security model, making NanoClaw a safer alternative to systems that rely on application-level security measures.
NanoClaw's philosophy centers around simplicity, customization, and AI-native design. The codebase is intentionally kept small and easy to understand, allowing users to modify and tailor the system to their specific requirements. Instead of relying on extensive configuration files, NanoClaw encourages users to customize their agents by directly modifying the code, leveraging the power of AI tools like Claude Code to assist in the process. This approach promotes a bespoke experience, where each user can create an AI assistant that perfectly matches their needs.
The system's architecture is designed to be AI-native, with the setup and onboarding process optimized for speed and efficiency. When judgment or customization is needed, control seamlessly shifts to Claude Code, allowing users to leverage AI to diagnose issues, make changes, and extend the system's functionality. NanoClaw supports scheduled tasks, allowing users to automate actions like sending daily briefings or updating information. It also provides web access, enabling agents to search and retrieve information from the internet.
NanoClaw's usage is straightforward. Users interact with their agents using a trigger word, such as "@Andy," followed by a command. For example, a user could instruct an agent to send a summary of the sales pipeline or update a document. The system also allows users to manage groups and tasks from within the messaging channels. Customization is achieved by instructing Claude Code to make specific changes to the code, such as modifying the trigger word or adding a custom greeting. The project encourages contributions through "skills" that add new channels or agent providers, keeping the core system minimal and allowing users to choose the features they need.
