Linkr KVM Server

Linkr KVM MCP Server exposes Radxa Linkr KVM controls through the Model Context Protocol (MCP). It lets an MCP client capture screenshots, crop screen areas, send keyboard and mouse events, type text, wait for on-screen text, and ask vision/OCR models about the current screen.

Requirements

• Python 3.11 or newer is recommended.
• A reachable Radxa Linkr KVM device.
• A Linkr API key.
• Vision/OCR provider API keys if using ask_vl or ocr_area.

Installation

Create and activate a virtual environment, then install dependencies:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

On POSIX-like shells:

python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt

Configuration

Copy the example configuration and fill in real credentials:

cp config.example.yaml config.yaml

Do not commit real secrets. Keep config.example.yaml sanitized.

Running

Recommended direct launch:

python main.py

The default transport is stdio. You can also select a transport explicitly:

python main.py --transport stdio
python main.py --http --host 127.0.0.1 --port 8000 --path /mcp
python main.py --transport sse --host 127.0.0.1 --port 8000 --sse-path /sse --message-path /messages/

Environment variables can also set defaults:

• MCP_TRANSPORT
• MCP_HOST
• MCP_PORT
• MCP_PATH
• MCP_SSE_PATH
• MCP_MESSAGE_PATH

The helper scripts forward arguments to the Python launcher:

.\start.ps1 --http

start.bat --transport stdio

./start.sh --http

MCP Tools

Tool	Description
snapshot	Capture the current KVM screen.
snapshot_area	Capture a selected screen area.
batch_events	Validate and send multiple Linkr KVM control events.
keyboard	Send keyboard press or release events using KeyboardEvent.code.
mouse_abs	Send absolute-position mouse events with normalized x and y.
mouse_rel	Send relative-position mouse events in pixels.
text	Type printable ASCII text, Tab, and Enter.
delay	Wait for a specified number of milliseconds.
ask_vl	Ask a vision-language model about a captured screen area.
ocr_area	Recognize text in a captured screen area with the OCR model.
find_text_pos	Find text positions on screen with local OCR.
wait_text_appear	Poll local OCR until text appears or a timeout is reached.

License

This project is licensed under the GPL-3.0-or-later License. See LICENSE for details.