Mercury Agent
Keeps a personal AI agent running on your computer that you can message from Telegram
by Cosmic Stack
QwenPaw
by AgentScope teamAgentGet organised
Gives you a personal AI assistant on your computer that you can message from chat apps
About QwenPaw
QwenPaw is an open source personal AI assistant you install on your own computer or server. You chat with it in a browser or apps like DingTalk and Telegram; it handles documents, web search and scheduled tasks.
For People who want a personal assistantTeams on DingTalk, Lark or DiscordLocal model users
Made by AgentScope team
A team that builds open source AI agent tools under the agentscope-ai GitHub organisation, including the AgentScope framework and the ReMe memory system that QwenPaw uses.
- Stars
- ★ 37k
- License
- Apache-2.0
- Cost
- Free
- Platforms
- Mac · Win · Linux
- Level
- Intermediate
Overview
QwenPaw is an open source personal AI assistant from the AgentScope team that you install on your own computer or server. You chat with it in a browser console, a terminal view or a beta desktop app, or from apps such as DingTalk, Lark, Discord, Telegram and QQ. It can work with documents, search the web and run tasks on a schedule, using a cloud model with your own API key or a local model that needs no key. It suits people who want one assistant that remembers their work and runs on their own machine.
It answers in the chat apps you already use, can run on local models with no API key, and keeps its memory as Markdown you can read and edit.
Try these first
Start with the first one: it is a small task you can check easily.
- Summarise this text in three bullet points: [paste text]First try · pasted text only
- Every weekday at 8am, send me a short digest of news about [topic].scheduled task
- Read [report].docx in my workspace and write a one-page summary next to it.writes a file
- Help me connect you to my [Discord] bot so I can message you there.connects a chat app
Example prompts written by Agent Flows Hub.
Features
From the creator's documentation.
| Feature | What it does | Runs |
|---|---|---|
| Console and desktop app | Chat in a browser console on your computer, a terminal view or a beta desktop app | Your computer |
| Chat app channels | Reach the same assistant from DingTalk, Lark, WeChat, Discord, Telegram, iMessage or QQ | Chat app accounts |
| Memory | Keeps conversations and notes as Markdown memory you can read, edit and search | Your computer |
| Scheduled tasks | Runs recurring jobs such as news digests and reports on your schedule | Your computer |
| Models | Cloud models with your API key, or local models through QwenPaw Local, Ollama or LM Studio | Model provider |
| Skills and plugins | Add skills for documents, browser and news, plus plugins and MCP tools | Other services |
Good to know
You want an assistant that lives on your own machine and replies in the chat apps you already use. QwenPaw runs a local console, connects to a cloud or local model, and can then take messages from channels such as DingTalk, Lark or Discord and run scheduled tasks.
Limits
- It does nothing until a model is set: add a provider API key under Settings, Models in the console, or download a local model. Local models need enough memory on your computer.
- Each chat app needs its own bot setup, and the desktop app is still in beta. If you deploy it on ModelScope Studio, set the Studio to non-public so others cannot control it.
- Running qwenpaw init --defaults also accepts anonymous install statistics (version, install method, operating system, Python version, CPU type and whether a GPU exists); run qwenpaw init without the flag to be asked.
Notes by Agent Flows Hub (original project name: QwenPaw), written from the creator's documentation, not from running it ourselves. QwenPaw quick start · QwenPaw README: install options and API keys
Stuck setting it up?
Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.
More to get organised
Other agents
- Rakazo Runs a team of AI assistants with their own jobs, memory and routines on your own setup
- Replit Builds and publishes apps and websites from a chat in plain words
- Rowboat Sorts your email, takes meeting notes and drafts replies using what it knows of your work
- Skales Runs as a desktop app that handles files, email, calendar and multi-step tasks for you
- Skyvern Fills in forms, downloads files and works through websites for you with an AI agent
- Tasklet Turns a task you describe in a chat into an agent that runs it across your apps