CoPaw is a lightweight and easy-to-use personal AI assistant. Its installation process is straightforward, and it can be deployed either on a local device or on a cloud server. CoPaw supports integration with multiple chat applications, offers strong extensibility, and ensures that your data remains under your control—never uploaded to third-party clouds. It proves useful in a variety of usage scenarios.
Summarize daily hot posts from platforms like Xiaohongshu, Zhihu, and Reddit. Generate content summaries for video platforms such as Bilibili and YouTube, enabling quick access to core information from various platforms.
Send summaries of various communications to office software like DingTalk, Feishu, and QQ. Extract contact information from emails and calendars, making the organization and transmission of office information more convenient.
Describe your creative goals to CoPaw, let it run in the background, and receive a first draft of your creation the next day, providing foundational content support for various creative tasks.
Track the latest news in technology and AI fields. Build your own knowledge base to quickly grasp industry trends and accumulate content related to personal learning and research.
Organize local files, read and summarize various document contents. Initiate file retrieval requests directly from the chat interface, making local file management and usage more efficient.
Combine skill functions with scheduled tasks to create your own intelligent applications. Tailor tools for different scenarios based on your usage habits and needs.
Suitable for users who prefer to manage their own Python environment. Execute the following three commands to complete the basic installation and startup:
pip install copaw
copaw init --defaults
copaw app
After installation and startup, open http://127.0.0.1:8088/ in your browser to access the CoPaw console. Here you can chat with CoPaw and configure your AI assistant. If you need to use it on platforms like DingTalk, Feishu, or QQ, refer to the official documentation for channel addition methods.
No need to manually install Python; the installation script handles the relevant environment configuration. Installation commands differ by system:
curl -fsSL https://copaw.agentscope.io/install.sh | bash
irm https://copaw.agentscope.io/install.ps1 | iex
After installation, open a new terminal and execute the following commands to start:
copaw init --defaults # Or run `copaw init` for interactive configuration
copaw app
# Install a specific version
curl -fsSL ... | bash -s -- --version 0.0.2
# Install from source (for development/testing)
curl -fsSL ... | bash -s -- --from-source
# Install version with local model support
bash install.sh --extras llamacpp # llama.cpp (cross-platform)
bash install.sh --extras mlx # MLX (Apple Silicon)
bash install.sh --extras llamacpp,mlx
# Upgrade by re-running the installation command
curl -fsSL ... | bash
# Uninstall
copaw uninstall # Keep configuration and data
copaw uninstall --purge # Delete everything
# Install a specific version
irm ... | iex; .\install.ps1 -Version 0.0.2
# Install from source (for development/testing)
.\install.ps1 -FromSource
# Install version with local model support
.\install.ps1 -Extras llamacpp # llama.cpp (cross-platform)
.\install.ps1 -Extras mlx # MLX (Apple Silicon)
.\install.ps1 -Extras llamacpp,mlx
# Upgrade by re-running the installation command
irm ... | iex
# Uninstall
copaw uninstall # Keep configuration and data
copaw uninstall --purge # Delete everything
# Pull the latest image
docker pull agentscope/copaw:latest
# Start the container
docker run -p 8088:8088 -v copaw-data:/app/working agentscope/copaw:latest
After startup, open http://127.0.0.1:8088/ in your browser to access the console. Configuration, memory, and skill data will be stored in the copaw-data volume. If you need to pass API keys (e.g., DASHSCOPE_API_KEY), add -e VAR=value or --env-file .env to the docker run command.
This image is built from scratch. If you need to build the image yourself, refer to the "Build Docker image" section in scripts/README.md, and then push it to your own image repository.
For users who prefer not to install locally, you can use ModelScope Studio for one-click cloud configuration. Remember to set your Studio to non-public status to prevent others from controlling your CoPaw.
To run CoPaw on Alibaba Cloud ECS, use the one-click deployment feature. Open the dedicated CoPaw deployment link on Alibaba Cloud ECS and follow the on-screen instructions. For detailed steps, refer to the Deploy Your AI Assistant in 3 Minutes article in the Alibaba Cloud Developer Community.
Suitable for development and customization. Execute the following commands:
# Clone the repository
git clone https://github.com/agentscope-ai/CoPaw.git
# Enter the project directory
cd CoPaw
# Install in editable mode
pip install -e .
# Install development dependencies (testing, formatting, etc.)
pip install -e ".[dev]"
# Build the console frontend
cd console && npm ci && npm run build
# Start the application from the project root
copaw app
If you use cloud-based large language models (e.g., Alibaba Cloud DashScope, ModelScope), you must configure the corresponding API key before chatting. CoPaw cannot function properly without a valid key.
copaw init, you will be prompted to configure the LLM provider and API key. Follow the terminal prompts to select a provider and enter your key.copaw app, open http://127.0.0.1:8088/ in your browser, go to Settings → Models, select the provider, fill in the API Key field, and activate the provider and model.DASHSCOPE_API_KEY in your shell or in a .env file in the working directory.Some tools require additional keys (e.g., TAVILY_API_KEY for web search). These can be configured in the console under Settings → Environment Variables. Refer to the project's configuration documentation for details.
If you only use local models (llama.cpp or MLX), no API key configuration is needed.
CoPaw supports running large language models locally, without relying on API keys or cloud services. Different backends suit different devices:
| Backend Framework | Suitable Device | Installation Command |
|---|---|---|
| llama.cpp | Cross-platform (macOS/Linux/Windows) | pip install 'copaw[llamacpp]' |
| MLX | Apple Silicon Macs (M1/M2/M3/M4) | pip install 'copaw[mlx]' |
copaw models download Qwen/Qwen3-4B-GGUF
copaw models
copaw app
You can also download and manage local models directly from the CoPaw console interface for a more intuitive experience.
CoPaw is developed by the AgentScope team, leveraging frameworks such as AgentScope, AgentScope Runtime, and ReMe.