Efflux desktop service
Efflux — An AI Copilot built for super-individuals.
Efflux Desktop integrates multi-vendor models and tool invocation capabilities, supporting command-based tool access and plugin extensions. Designed for real-world tasks, Efflux brings human-AI collaboration back under your control — amplifying the judgment and execution power of super-individuals.
It doesn’t decide for you — it helps you stay in control.
Multi-vendor AI model integration (OpenAI, Anthropic, DeepSeek, etc.)
Natural-language-based conversations
Text-to-artifact capabilities
Real-time streaming chat responses
Chat history management
Dynamic discovery and loading of MCP servers
Tool configuration management support
Exception handling and timeout control
Standardized tool calling interface
Download the package (e.g., efflux-desktop-mac-arm.zip for macOS) to your local disk.
Unzip the package and double-click the application file (e.g., EFFLUX-1.0.0-x64.dmg) to install Efflux Desktop.
Unzip the executable file (e.g., efflux_desktop) and double-click it to load the required services.
Note that this extra executable file is only required at this moment and will be eliminated in future releases.
In the navigation pane of Efflux Desktop, select Models.
In the Model Providers page, find your desired model vendor, and click API-KEY.
In the pop-up dialog, enter your endpoint and API key, and click Save.
Optional: If you want to use existing MCP servers to complete your task, do the following:
In the navigation pane, select Plugins.
In the Discover Plugins tab, click Add Custom Plugin.
In the pop-up dialog, do either of the following and click Add.
Enter the plugin name, command, environment variables, and arguments (if any), or
If you've already got a JSON string, select JSON Mode and paste it.
Tip
You can find the JSON resources of MCP servers from the following websites.
In the navigation pane, select Chat.
In the chatbox, select the model you've configured, and:
Tell Efflux your question, or
Switch to the Build mode and describe what you want Efflux to build.
To use the installed plugin, enter the @ sign and select the target one.
Press the Enter key to start your conversation with the selected model.
efflux-desktop/
├── adapter/ # Adapter layer
│ ├── mcp/ # MCP protocol adapters
│ ├── model_sdk/ # Model SDK adapters
│ ├── persistent/ # Persistence adapters
│ └── web/ # Web interface adapters
├── application/ # Application layer
│ ├── domain/ # Domain objects
│ ├── port/ # Port interfaces
│ └── service/ # Application services
├── common/ # Common components
│ ├── core/ # Core infrastructure
│ └── utils/ # Utility classes
└── main.py # Application entry pointNote that this section is for developers who are interested in contributing to this project.
git clone https://github.com/isoftstone-data-intelligence-ai/efflux-desktop.git
cd efflux-desktopInstall dependencies using the uv package manager:
pip install uv
uv sync --reinstallActivate a virtual environment and configure environment variables.
# Activate virtual environment
source .venv/bin/activate # MacOS/Linux
# Deactivate when needed
deactivateuv runThe service will start at http://127.0.0.1:8000.
POST /api/agent/chat/default_chat
Content-Type: application/json
{
"firm": "openai",
"model": "gpt-4",
"system": "You are a helpful AI assistant",
"query": "Hello, please introduce yourself",
"mcp_name_list": ["example-server"]
}Fork this project.
Create a feature branch (git checkout -b feature/AmazingFeature).
Commit your changes (git commit -m 'Add some AmazingFeature').
Push to the branch (git push origin feature/AmazingFeature).
Submit a Pull Request.
This project follows the appropriate open source license. Please refer to the LICENSE file for details.
For questions or suggestions, please contact us through:
Submit Issues
Start Discussions
No open issues yet, or sync has not completed.