zofrasca/lightclaw

LightClaw：真正的轻量级AI助手

LightClaw 是一款快速、本地优先的人工智能助手，灵感来自 OpenClaw 和纳米机器人，以单一的 Rust 二进制形式包装。

如果你想要代理工具、内存和Telegram/Discord集成，但运行时间不长，Lightclaw就是为此设计的。

为什么是光爪

60秒快速入门

支持的平台：

• Linux x86_64
• Linux ARM64（树莓派4/5，ARM服务器）
• Linux ARMv7（例如较旧的ARM设备/电子阅读器）
• macOS x86_64
• macOS ARM64（苹果硅片）

注意：存在Windows二进制文件，但目前稳定性较差，支持也不如其他二进制文件。

你得到了什么

• 单二进制部署：提供一个可执行文件，不包含 Python 运行时。
• 具备工具支持的代理：文件、壳、网页和调度操作。
• Telegram/Discord原生接口：内置高性能轮询功能。

LightClaw is a fast, local-first AI assistant inspired by OpenClaw and nanobot, packaged as a single Rust binary.

If you want agentic tooling, memory, and Telegram/Discord integration without a heavy runtime, lightclaw is built for that.

Why lightclaw

60-Second Quickstart

curl -fsSL https://lightclaw.dev/install.sh | bash
lightclaw configure

Supported platforms:

• Linux x86_64
• Linux ARM64 (Raspberry Pi 4/5, ARM servers)
• Linux ARMv7 (e.g. older ARM devices / e-readers)
• macOS x86_64
• macOS ARM64 (Apple Silicon)

Note: A Windows binary exists, but it is currently less stable and not as well-supported as the others.

What You Get

• Single-binary deploy: ship one executable, no Python runtime.
• Tool-capable agent: file, shell, web, and scheduling actions.
• Telegram/Discord-native interface: high-performance polling built in.
• Local-first memory: vectors + metadata stored locally with SQLite.
• Rust reliability: strong typing, memory safety, and concurrency.
• Skills support: OpenClaw-style SKILL.md skills via activate_skill.

Memory System

LightClaw includes long-term memory:

• Short-term chat history per session.
• Periodic summarization of recent conversation chunks.
• Semantic retrieval over stored memories.
• Privacy-first local storage (no external vector DB required).

Configuration

Create ~/.lightclaw/config.json:

{
  "agents": {
    "defaults": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4-5",
      "model_fallbacks": [
        "openai/gpt-4o-mini",
        "ollama/llama3.2"
      ]
    }
  },
  "providers": {
    "openrouter": {
      "apiKey": "sk-or-..."
    },
    "openai": {
      "apiKey": "sk-..."
    },
    "ollama": {
      "apiBase": "http://127.0.0.1:11434/v1"
    },
    "mistral": {
      "apiKey": "..."
    }
  },
  "channels": {
    "telegram": {
      "token": "YOUR_BOT_TOKEN",
      "allow_from": ["123456789"],
      "transcription": {
        "enabled": true,
        "provider": "openai",
        "model": "whisper-1",
        "language": "en",