ABOUT

QwenPaw is a personal AI assistant and Agent OS built by agentscope-ai. Running locally or in the cloud, it utilizes fine-tuned QwenPaw-Flash models (2B/4B/9B) alongside local execution runtimes like Ollama and LM Studio. It features a robust three-layer memory system (live, historical, and distilled) that retains context without losing old details. QwenPaw prioritizes security with kernel-level sandboxing and guardrails, offers a 3-panel Web IDE Coding Mode, supports multi-agent parallel workflows via the Agent Communication Protocol, and seamlessly integrates with platforms like WeChat, DingTalk, Lark, and Discord. Highly extensible, it supports MCP, plugins, and custom skills via its dedicated hub.

CAPABILITIES

Three-layer Persistent Memory
Local & Cloud Hybrid Deployment
Kernel-level Sandbox & Security Guard
Three-panel Web IDE Coding Mode
Multi-Agent Orchestration & Multi-Channel Integration

SUPPORTED PLATFORMS

webdesktopwechatmobile

EXTERNAL RESOURCES

Visit Website ↗ GitHub Repository ↗ Skills Market ↗

RELATED ECOSYSTEM PRODUCTS