What is QwenPaw used for?

QwenPaw is a personal AI assistant workstation you run locally or in the cloud, with layered memory, a plugin and skill system, MCP support and many chat-app channels. It is a good fit for a personal assistant reachable from messaging apps, private AI workflows with local models and automating recurring tasks with agent skills.

Is QwenPaw open source?

Yes. QwenPaw is open source under the Apache-2.0 licence. Its source code is on GitHub at agentscope-ai/QwenPaw and is written mainly in TypeScript.

Yes. QwenPaw is open source, so the software itself is free to use.

Can I self-host QwenPaw?

Yes. QwenPaw can be self-hosted on your own server or infrastructure; there is no official hosted version.

What are some alternatives to QwenPaw?

Similar open-source tools in the AI Tools category include Nanobot, LobeHub and AnythingLLM. SaaS products in the same category include Aera Browser, AlfredClaw and BrainSite.

Is QwenPaw actively maintained?

Yes. The most recent commit to QwenPaw was on 30 September 2026, and the latest release is v2.2.1, published on 11 September 2026. The project has 35k stars on GitHub.

QwenPaw: open-source AI tools

About QwenPaw

QwenPaw is an open-source personal AI agent workstation that runs in your own environment, either on your machine or in the cloud. It is easy to install and talks to you through multiple chat apps as well as a console, a terminal UI and a desktop app, so one instance can follow you across channels.

Memory has three layers: the live working context, a complete word-for-word history, and a personal knowledge base that keeps evolving as conversations and resources are turned into readable, editable Markdown notes. Models can be local, using the QwenPaw-Flash family (2B, 4B and 9B) with a built-in local runtime and no API key, or connected through Ollama, LM Studio or more than a dozen cloud providers. Safety features include a kernel-level sandbox, tool and file guards, a skill scanner and access policies. It also supports independent sub-agents, the Agent Communication Protocol, a file workspace and MCP integration, with channels such as DingTalk, Lark, WeChat, Discord, Telegram, iMessage and QQ.

QwenPaw comes from the AgentScope team, is Apache-2.0 licensed and is written in TypeScript. Skills for scheduling, documents, browsing and news, plus a plugin marketplace, extend it. Data stays on your machine when deployed locally.

Key features

Three-layer memory stored as Markdown
Local models or 14+ cloud providers
Sandbox, tool guard and access policies
Multi-agent and parallel sub-agents
Skills, plugins and MCP integration
Chat channels including Discord, Telegram and WeChat

Good fit for

→A personal assistant reachable from messaging apps
→Private AI workflows with local models
→Automating recurring tasks with agent skills

Built with: TypeScript
Tags: ai-agent; personal-assistant; llm; mcp; self-hosted; multi-agent; local-ai; agentscope; chatbot

QwenPaw: questions and answers

What is QwenPaw used for?: QwenPaw is a personal AI assistant workstation you run locally or in the cloud, with layered memory, a plugin and skill system, MCP support and many chat-app channels. It is a good fit for a personal assistant reachable from messaging apps, private AI workflows with local models and automating recurring tasks with agent skills.
Is QwenPaw open source?: Yes. QwenPaw is open source under the Apache-2.0 licence. Its source code is on GitHub at agentscope-ai/QwenPaw and is written mainly in TypeScript.
Is QwenPaw free?: Yes. QwenPaw is open source, so the software itself is free to use.
Can I self-host QwenPaw?: Yes. QwenPaw can be self-hosted on your own server or infrastructure; there is no official hosted version.
What are some alternatives to QwenPaw?: Similar open-source tools in the AI Tools category include Nanobot, LobeHub and AnythingLLM. SaaS products in the same category include Aera Browser, AlfredClaw and BrainSite.
Is QwenPaw actively maintained?: Yes. The most recent commit to QwenPaw was on 30 September 2026, and the latest release is v2.2.1, published on 11 September 2026. The project has 35k stars on GitHub.