Quick Start

Option A: Script install

The script install is the fastest path for most users:

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

Why this path exists:

• it avoids manual Python setup;
• it can bootstrap the environment with uv;
• it is friendlier for first-time users who want the workstation running quickly.

After installation, open a new terminal and run:

copaw init --defaults
copaw app

The browser Console usually opens on http://127.0.0.1:8088/.

Option B: pip install

If you prefer to manage Python environments yourself:

pip install copaw
copaw init --defaults
copaw app

This path is a good fit when:

• your team already uses reproducible Python environments;
• you want tighter control over the installation process;
• you are testing changes against a local Python workflow.

Option C: Docker

For explicit infra and persistent data:

docker pull agentscope/copaw:latest
docker run -p 8088:8088 -v copaw-data:/app/working agentscope/copaw:latest

Docker is useful when you want:

• a stable self-hosted runtime;
• mounted or named-volume persistence;
• a deployment story that is easier to move from laptop to server.

Option D: Cloud-oriented setup

CoPaw also supports cloud-assisted paths such as ModelScope-based flows. That is the right direction when:

• you do not want to keep the first runtime on your own laptop;
• you want a fast hosted evaluation path;
• you still care about the product model, but not the first install friction.

Verify the install

After the first launch, confirm four things:

1. the Console loads;
2. you can configure a provider;
3. you can send a test chat;
4. your working directory exists and persists.

When to route to EasyClaw

If the question becomes "How do I take this beyond a local evaluation and into a traffic-ready production deployment?", use the deployment handoff path to EasyClaw. This site treats that as the proper operational CTA rather than pretending documentation alone covers the full rollout journey.