mcp-serial-hid-kvm
Gives AI agents full keyboard, mouse, and screen access to a physical PC via a KVM server and OCR.
README
mcp-serial-hid-kvm
MCP (Model Context Protocol) server that gives AI agents full keyboard, mouse, and screen access to a physical PC. Thin client for serial-hid-kvm — all hardware control is delegated via TCP.
How It Works
Claude / AI Agent
↕ MCP (stdio)
mcp-serial-hid-kvm ← this package (thin client + OCR)
↕ TCP (localhost:9329)
serial-hid-kvm ← standalone KVM server (owns hardware)
↕ USB Serial + HDMI
Target PC
The KVM server (serial-hid-kvm) runs as a persistent process owning the serial port and capture device. This MCP server connects to it as a TCP client. Multiple MCP instances (multiple Claude sessions) can share a single KVM server without device conflicts.
Prerequisites
- Hardware: CH9329+CH340 USB HID cable + USB HDMI capture device (see serial-hid-kvm for details)
- serial-hid-kvm installed and running:
pip install -e /path/to/serial-hid-kvm serial-hid-kvm --api # with preview window serial-hid-kvm --api --headless # or headless - Tesseract OCR (for
get_screen_text/execute_and_read):- Linux:
sudo apt install tesseract-ocr - Windows: https://github.com/tesseract-ocr/tesseract
- Linux:
Installation
pip install -e .
This automatically installs serial-hid-kvm as a dependency.
MCP Client Configuration
Claude Desktop / Claude Code
{
"mcpServers": {
"kvm": {
"command": "mcp-serial-hid-kvm"
}
}
}
Custom KVM server address:
{
"mcpServers": {
"kvm": {
"command": "mcp-serial-hid-kvm",
"env": {
"SHKVM_API_HOST": "127.0.0.1",
"SHKVM_API_PORT": "9329"
}
}
}
}
Configuration
| Variable | Default | Description |
|---|---|---|
SHKVM_API_HOST |
127.0.0.1 |
KVM server address |
SHKVM_API_PORT |
9329 |
KVM server port |
SHKVM_OCR_CMD |
auto-detect | Path to tesseract executable |
SHKVM_CAPTURE_LOG_DIR |
platform default | Capture log directory (empty string to disable) |
Hardware settings (SHKVM_SERIAL_PORT, SHKVM_SCREEN_WIDTH, etc.) are configured on the KVM server side, not here. If the target PC uses a non-US keyboard, set --target-layout (or SHKVM_TARGET_LAYOUT) on the KVM server so that type_text and send_key produce correct characters.
Available Tools
Keyboard
| Tool | Description |
|---|---|
type_text |
Type text with inline tags: ls -la{enter}, {ctrl+c}, {alt+f4}. Whitelist-based: unknown {content} passes through literally. Raw mode (raw=true) disables tags; actual line breaks become Enter. char_delay_ms: delay between keystrokes in ms (default: 20). Only ASCII printable characters, tab, and line breaks are supported; unsupported characters (Unicode, CJK, etc.) cause an error — use base64 encoding as a workaround |
send_key |
Single key press with modifiers |
send_key_sequence |
Multiple key steps with per-step delays. default_delay_ms: delay between steps in ms (default: 100); each step can override with delay_ms |
Mouse
| Tool | Description |
|---|---|
mouse_move |
Move cursor (absolute or relative) |
mouse_click |
Click at optional position |
mouse_drag |
Drag from one position to another (drag-and-drop, text selection, etc.) |
mouse_scroll |
Scroll wheel |
Screen
| Tool | Description |
|---|---|
capture_screen |
Capture screen as image (high token cost) |
get_screen_text |
Capture + OCR to text (preferred for text content) |
execute_and_read |
Type command, Enter, wait, capture + OCR |
Device Management
| Tool | Description |
|---|---|
get_device_info |
Serial port, capture device, config info |
list_capture_devices |
List available video devices |
set_capture_device |
Switch capture device |
set_capture_resolution |
Change capture resolution |
Architecture
This package is intentionally minimal (~4 files):
mcp_serial_hid_kvm/
server.py MCP tool handlers → KvmClient TCP calls
config.py KVM host/port, tesseract, log settings
ocr.py Tesseract OCR (runs locally on fetched frames)
__init__.py
All keyboard/mouse/capture logic lives in serial-hid-kvm. This package only translates MCP tool calls to TCP API calls and runs OCR locally.
Why Separate?
- No device conflicts — multiple Claude sessions share one KVM server
- Independent restarts — restart the MCP server without losing the KVM connection
- Standalone use —
serial-hid-kvmworks without MCP (interactive preview, scripts, other AI frameworks)
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.