Run a Personal AI Assistant Locally with QwenPaw

// Before you start

What you need

A computer running macOS, Linux, or Windows
Internet connection for the initial download
An API key from a provider like OpenAI, Anthropic, or Qwen (optional — local models work too)
About 2 GB of free disk space

Step 1 of 5

Install QwenPaw with one command

5 min

This single command downloads a setup script that handles everything: it installs a Python environment manager called uv, creates an isolated environment, and installs QwenPaw with all its dependencies including the web frontend. You do not need Python pre-installed for this path.

Terminal · mac

$ macOS / Linux:

$ curl -fsSL https://qwenpaw.agentscope.io/install.sh | bash

$ Windows CMD:

$ curl -fsSL https://qwenpaw.agentscope.io/install.bat -o install.bat && install.bat

$ Windows PowerShell:

$ irm https://qwenpaw.agentscope.io/install.ps1 | iex

What you should see

A series of download and install messages ending with something like 'QwenPaw installed successfully' and instructions to run qwenpaw.

This might happen

The script fails or hangs — common on corporate networks or VPNs with strict firewalls.

Switch to the pip install method instead: first install Python 3.11–3.13 from python.org, then run: pip install qwenpaw

Step 2 of 5

Initialise QwenPaw with default settings

1 min

This command creates the configuration folder and default settings file in your home directory. It only needs to be run once. The --defaults flag skips interactive prompts so you can get started immediately.

Terminal · mac

$ qwenpaw init --defaults

What you should see

A message like 'Initialized QwenPaw at ~/.qwenpaw' with no errors.

This might happen

'qwenpaw' is not recognised as a command on Windows after the script install.

Close your terminal completely and reopen it so the new PATH entry takes effect. If it still fails, manually add %USERPROFILE%\.qwenpaw\bin to your system PATH via System Properties > Environment Variables.

Step 3 of 5

Start the QwenPaw app

1 min

This launches the QwenPaw backend server and opens the web interface. Once it starts, your browser is your control panel for everything — adding a model, chatting, and configuring channels.

Terminal · mac

$ qwenpaw app

What you should see

Terminal shows 'Server running at http://127.0.0.1:8088/' or similar. Your default browser may open automatically.

This might happen

Port 8088 is already in use by another application.

Stop the other application using that port, or check QwenPaw's documentation for a flag to change the port.

Step 4 of 5

Open the browser console and add a model

5 min

Navigate to the address QwenPaw printed in your terminal. The first thing you need to do is tell QwenPaw which AI model to use. Go to the Models page in the sidebar. You can paste in an API key from a cloud provider (OpenAI, Anthropic, Qwen, Gemini, and 14+ others are supported), or point it at a locally running model via Ollama or LM Studio if you have one set up.

Terminal · mac

$ Open in your browser:

$ http://127.0.0.1:8088/

What you should see

The QwenPaw web interface loads. The sidebar shows sections including Models, Sessions, and Skills.

This might happen

The page shows a blank screen or fails to load.

Wait 15–20 seconds and refresh — the frontend assets may still be compiling on first launch. If it persists, check the terminal for error messages.

Step 5 of 5

Start a chat session

2 min

Once a model is configured, click 'New Session' (or the equivalent button) in the sidebar to open a chat window. Type a message and press Enter. QwenPaw stores the full conversation history and can recall older turns even after they leave the active context window.

Terminal · mac

$ No command needed — use the web interface at http://127.0.0.1:8088/

What you should see

The assistant replies in the chat window. Token usage appears per turn.

// Status

cooked. baked. worked.

A personal AI assistant running entirely on your own machine, accessible at http://127.0.0.1:8088/, with persistent conversation memory and the ability to connect to a cloud or local AI model of your choice.

// the honest bit

The honest part

QwenPaw is a fast-moving project — the README lists many features (multi-agent, Coding Mode, channel integrations, Skills marketplace) that require additional configuration steps not covered here. This guide gets you to a working local chat only. Connecting to DingTalk, Lark, Discord, or Telegram requires reading the Channel setup docs separately. Local model support via Ollama requires Ollama to already be installed and running. The free QwenPaw-Flash models (2B–9B) need a capable machine — a modern laptop with 8 GB RAM is the minimum for the smaller sizes.