QwenPaw — Self-Hosted Personal AI Assistant with Multi-App Support

Introduction

QwenPaw is an open-source, self-hosted personal AI assistant framework. It provides a web UI and extensible backend that connects to multiple LLM providers and chat applications. Built by the AgentScope team, it emphasizes easy deployment on your own machine or cloud server with full control over your data and model choices.

What QwenPaw Does

• Provides a web-based chat interface for interacting with AI models
• Connects to multiple chat platforms for unified assistant access
• Supports MCP tools for extending agent capabilities
• Integrates with various LLM providers including local models
• Manages conversation history and context across sessions

Architecture Overview

QwenPaw is a Python application built on a modular agent framework. The core consists of a conversation manager, a tool execution layer supporting MCP servers, and a multi-provider LLM backend. The web UI is served via Gradio, and chat platform connectors run as separate workers that bridge external messaging services to the central agent loop.

Self-Hosting & Configuration

• Requires Python 3.10+ and pip for installation
• Configure LLM provider endpoints and API keys in settings
• Optionally connect chat platform bots via their respective API tokens
• Deploy with Docker for a containerized setup
• Supports GPU acceleration for local model inference

Key Features

• Multi-provider LLM support including Qwen, OpenAI, and Ollama
• Extensible tool system via Model Context Protocol
• Web UI with conversation history and session management
• Plugin architecture for adding new chat platform connectors
• Self-hosted with all data stored locally

Comparison with Similar Tools

• Open WebUI — focuses on Ollama/local models, QwenPaw adds multi-platform chat connectors
• LibreChat — multi-provider chat UI without the agent tool framework
• LobeChat — polished UI with plugin marketplace, less focus on self-hosted agent workflows
• Jan — desktop-first offline AI app without multi-platform chat bridging
• Chatbot UI — minimal ChatGPT-style interface without tool or MCP support

FAQ

Q: Which LLM providers are supported? A: QwenPaw supports Qwen models, OpenAI-compatible APIs, Anthropic, Google, and local models through Ollama.

Q: Can I run it entirely offline? A: Yes, by configuring a local model provider like Ollama, you can run QwenPaw without any external API calls.

Q: What chat platforms can it connect to? A: It supports several popular messaging platforms through its connector plugin system. Check the repository for the current list.

Q: Does it support function calling and tools? A: Yes, it has built-in MCP tool support and can execute functions defined in connected MCP servers.

Sources

• https://github.com/agentscope-ai/QwenPaw
• https://github.com/agentscope-ai/QwenPaw#readme