nanocoai/nanoclaw

A lightweight alternative to OpenClaw that runs in containers for security. Connects to WhatsApp, Telegram, Slack, Discord, Gmail and other messaging apps,, has memory, scheduled jobs, and runs directly on Anthropic's Agents SDK

What it solves

NanoClaw 是一個輕量級的 AI 助手框架,專為想要在多個訊息平台上以安全、可自訂且保護隱私的方式執行 AI 代理的使用者設計。它透過將代理隔離在各自的 Linux 容器中,避免了執行具有廣泛系統存取權限的複雜 AI 軟體所帶來的安全風險,確保代理只能存取明確掛載的目錄。

How it works

NanoClaw 使用單一的 Node.js 主機程序作為路由器。當來自訊息應用(如 Telegram 或 Discord)的訊息到達時,主機會將其路由至特定代理群組的會話,並寫入 inbound SQLite 資料庫。執行 Claude Agent SDK(透過 Bun)的沙盒 Docker 容器會輪詢此資料庫,處理請求,並將回應寫入 outbound SQLite 資料庫,之後主機再將回應傳回給使用者。

Who it’s for

此框架適合希望擁有最小化、"AI 原生" 程式碼基礎的個人使用者,他們可以輕鬆理解、fork 並使用 Claude Code 等 AI 工具自行客製化,而不必依賴大型、單塊式的框架與複雜的設定檔。

Highlights

  • Container Isolation: Agents run in Docker containers to prevent host system access.
  • Multi-channel Support: On-demand installation of adapters for WhatsApp, Telegram, Discord, Slack, and more.
  • Credential Security: API keys are managed via OneCLI's Agent Vault, so agents never hold raw keys.
  • Bespoke Customization: Designed to be modified directly in the code via AI prompts rather than through config files.
  • Scheduled Tasks: Ability to set up recurring jobs with optional script gates.
  • Flexible Model Providers: Native support for Claude, with options for OpenAI, Google, DeepSeek, and local models via Ollama.