NanoClaw is a lightweight, open-source personal AI agent system designed to let you run and control AI assistants that can talk, act, and automate tasks across your apps. It runs on your own machine and uses container isolation (Docker or Apple Containers) so each agent is sandboxed at the operating system level for security. You interact with it through messaging apps like WhatsApp, Telegram, Discord, Slack, and email, turning chat into a control layer for your AI.

• Container-Isolated Security: Every agent session runs in its own Docker container with full filesystem isolation enforced by the OS, not application code. Even if a session is compromised, it can't affect others.
• Multi-Channel Messaging: Connect agents to messaging platforms and services including WhatsApp, Telegram, Discord, Slack, Microsoft Teams, email, and more through installable skills.
• Agent Swarms: Create teams of specialized AI agents that can work together on larger tasks and parallel workflows.
• Scheduled Tasks & Memory: Automate daily reports, reminders, and background jobs with a built-in task scheduler. SQLite-based memory gives each group its own persistent context.
• Minimal, Auditable Codebase: The entire project fits in ~15 files and ~700 lines of TypeScript. You can read, understand, and modify every line of your agent's code.
• Built-In Web Browsing: The container agent can open pages, fill forms, take screenshots, and extract data directly from the web.
• Customizable Workflow Model: IFork and tailor your AI assistant to match your own workflow and requirements.
• Multiple AI Providers: Runs natively on Claude via Anthropic's Agent SDK, with drop-in support for OpenAI Codex, OpenRouter, Google, DeepSeek, and local models via Ollama.

• Strong Security by Default: OS-level container isolation is a real security boundary, not a software sandbox. Each messaging group gets its own isolated container, memory, and filesystem.
• Lightweight and Easy to Audit: NanoClaw is designed with a smaller codebase and simple architecture, making it easier to understand and customize than larger AI agent platforms.
• Fast Setup: The project includes a streamlined installer that handles setup steps and helps users get started quickly.
• Practical Multi-Channel Support: One agent can handle WhatsApp, Telegram, Slack, Discord, and email simultaneously with unified memory across channels.
• Active Community & Development: Praised by notable figures like Andrej Karpathy for its manageability and auditability. Open source under MIT license.

• Requires Technical Comfort: You need familiarity with the terminal, Git, and Docker. There's no GUI or dashboard for setup or management.
• Claude-Optimized Workflow: Claude and the Claude Agent SDK are the default experience, although additional model providers can be connected.
• Fewer Built-In Integrations: Compared to OpenClaw, NanoClaw ships with fewer out-of-the-box integrations. Channels and skills must be added individually.
• Container Runtime Dependency: NanoClaw relies on Docker or Apple Container environments for its security and isolation model, which may require additional system resources and initial setup steps.

• Personal AI Assistant via Messaging: Chat with your AI agent through WhatsApp or Telegram just like you would a human assistant. Send messages back-to-back, forward emails and notes, and get organized responses with follow-up reminders throughout your day.
• Business Operations & Sales Management: Turn messy messages, emails, and notes into structured actions. Agents can generate daily or scheduled briefings, summarize updates, and create follow-up tasks using the built-in scheduling system and persistent memory per messaging group.
• DevOps & Engineering Workflows: Connect GitHub, Jira, and APIs to generate engineering summaries, release reports, and workflow updates. Each agent runs in an isolated container, helping keep different projects or clients separated while maintaining consistent automation pipelines.
• Research & Documentation Automation: Schedule weekly tasks to review git history, summarize changes, and generate wiki-ready documentation. Researchers on shared lab machines benefit from OS-level isolation keeping their notes and data separate.
• Web Automation & Data Extraction: Use built-in browser capabilities inside containers to navigate pages, extract data, and interact with web interfaces. This enables automation for research, monitoring, and structured data collection workflows.
• Multi-Agent Collaboration: Break complex tasks into smaller parts handled by multiple specialized agents. One agent can gather data, another can analyze it, and another can write the final output. These coordinated workflows run through agent-to-agent orchestration inside the system.