Coordinator

You use the MCP (Model Context Protocol) to coordinate and access multiple MCP servers from a central coordinator. This lets you perform KDE-related operations, secure code execution, data processing, and network tasks through a unified interface, improving automation and cross-component collaboration.

How to use

To use MCP you connect your client to the central coordinator at the provided URL and then invoke the available MCP operations through the client. You can query server capabilities, perform KDE actions like KRunner searches and file operations, execute code securely, analyze data, visualize results, and make network requests or downloads. Use the coordinator as the single entry point to access all MCP servers.

How to install

Prerequisites you must have before using MCP features are Node.js (v14 or later) and the MCP servers running, plus required Python packages for data visualization.

Step 1. Install the MCP Servers

cd ~/GIT-Projects/MCP-Servers
npm install

Step 2. Start the MCP Servers

npm run start:all

Alternatively, you can start all MCP servers using the Python launcher script included in the KDE AI Interface

cd ~/GIT-Projects/KDE\ AI\ Interface
./launch_mcp_servers.py --start

Step 3. Enable MCP in your interface

from app_root.config.settings import SettingsManager
s = SettingsManager()
print(f"MCP enabled: {s.get('mcp/enabled')}")

If MCP is not enabled, enable it and save the settings

s.set('mcp/enabled', True)
s.save()

Additional setup notes

If you want to verify the MCP client setup and test operations, you can run the test client to exercise the MCP integration. The test client opens a GUI with tabs for regular LLM chat and MCP operations.

cd ~/GIT-Projects/KDE\ AI\ Interface
source venv/bin/activate
python test_mcp_client.py

Available tools

get_capabilities

Retrieve capabilities from all MCP servers.

krunner_query

Query KDE KRunner with a search term.

list_home_directory

List files in the user home directory.

read_file

Read a file from a given path.

send_notification

Send a desktop notification through KDE.

execute_code

Execute code on the Code Execution MCP Server.

analyze_data

Analyze data via the Data Processing MCP Server.

plot_data

Create a plot from analyzed data.

http_request

Make a network HTTP request via the Network Operations MCP Server.

download_file

Download a file via the Network Operations MCP Server.