lg-tv-mcp
Turns AI assistants like Claude and ChatGPT into a remote control for LG webOS smart TVs, enabling power control, volume, app launching, input switching, and more, all locally without cloud APIs.
README
lg-tv-mcp
עברית: מדריך למשתמש בעברית בהמשך העמוד ומדריך התקנה מלא ב־docs/client-installation-he.md.
A local MCP server that turns an AI assistant — Claude Code, Claude Desktop, ChatGPT Work, or any MCP client — into a remote control for LG webOS smart TVs.
Everything runs on your home network: a hand-rolled SSAP websocket client with first-run pairing, the webOS button-input socket, Wake-on-LAN power-on, and mDNS re-discovery when the TV's IP changes. No cloud APIs, no accounts, no telemetry.
Tools
| Tool | What it does |
|---|---|
get_status |
Power, pairing state, current app, volume — no side effects |
power_on |
Wake-on-LAN magic packet; TV boots in ~15 s |
power_off |
Turn the TV off over SSAP |
set_volume |
Set level 0–100, or mute/unmute |
press_button |
Remote buttons (arrows, OK, back, play/pause, digits…), with optional repeat |
list_apps / launch_app |
Enumerate and open apps (Netflix, YouTube, …) |
switch_input |
Switch HDMI inputs |
set_channel |
Change Live TV channels |
type_text |
Type into the focused text field |
show_notification |
Toast message on the TV screen |
screenshot |
960×540 JPEG of the current screen (DRM-protected video captures as black; app UI renders fine) |
measure_sound_level |
Room loudness via the laptop microphone: approximate dB SPL + band (silent/quiet/moderate/loud/very-loud). Loop it with set_volume to hit a target level |
calibrate_sound_meter |
One-time mic calibration against a phone SPL-meter reading; the offset persists in config.json |
Requirements
- Node.js ≥ 22.13
- An LG webOS TV on the same network as the machine running the server
- For powering on while the TV is off: enable Settings → General → Turn on via Wi-Fi on the TV (wording varies by webOS version), and set your TV's MAC address in the config — see Configuration
ffmpegfor the sound-meter tools (brew install ffmpeg); macOS microphone permission is requested from the hosting app on first capture.
Install
Not yet on npm — install from a clone:
git clone https://github.com/OmerLapidot/lg-tv-mcp.git
cd lg-tv-mcp
npm ci
Claude Code
claude mcp add lg-tv-remote --scope user -- node /absolute/path/to/lg-tv-mcp/server/index.js
Claude Desktop / any MCP client
{
"mcpServers": {
"lg-tv-remote": {
"command": "node",
"args": ["/absolute/path/to/lg-tv-mcp/server/index.js"]
}
}
}
ChatGPT Work (desktop plugin)
npm run plugin:release builds dist/plugin/ with the plugin manifest, skill, and icon; the release runtime pins npx -y @omerlapidot/lg-tv-mcp@<version> (npm publish is a separate release step). Installation walkthrough (Hebrew): docs/client-installation-he.md.
First run — pairing
- Turn the TV on with the physical remote (first time only) and ask the assistant something harmless like "what's the TV status?" —
get_statusworks unpaired. - The first control call (volume, buttons, apps…) pops a pairing prompt on the TV screen. Accept it with the physical remote within a minute.
- That's it — the TV issues a client key, saved locally. The prompt never appears again unless you revoke the pairing on the TV (Settings → General → External Devices → Connection Manager).
Configuration
State lives in a single JSON file (created and updated automatically):
- macOS:
~/Library/Application Support/lg-tv-mcp/config.json - Windows:
%APPDATA%\lg-tv-mcp\config.json
| Field | Meaning |
|---|---|
hostname |
mDNS name tried when the last-known IP stops answering. LGwebOSTV.local is the LG factory default, so this usually just works. |
ip |
Last-known IPv4 of the TV. Auto-updated whenever discovery finds the TV at a new address. |
mac |
The TV's Wi-Fi MAC, used only for Wake-on-LAN. The repo default is a placeholder — set your own (shown on the TV under Settings → Network → Wi-Fi Connection → Advanced) or power_on cannot wake a TV that is fully off. |
clientKey |
Pairing key issued by the TV on first approval; null until paired. |
micCalibrationOffsetDb |
Optional. dBFS→dB SPL offset stored by calibrate_sound_meter; absent until first calibration (a rough default of +100 is used, flagged calibrated: false). |
Environment overrides: LG_TV_DATA_DIR (config directory), LG_TV_URL (skip discovery and connect to a fixed ws:///wss:// URL — used to point at the fake TV in tests), LG_TV_REGISTER_TIMEOUT_MS (pairing prompt timeout).
Security notes
- The client key works only from inside your LAN and can only control this TV; it contains no account credentials.
- The TV serves SSAP and screen captures over TLS with a self-signed certificate, so certificate validation is intentionally disabled for those LAN-only connections.
- Nothing leaves your network; there is no cloud relay.
Development
npm ci
npm test # full suite against an in-process fake TV — no hardware needed
npm run plugin:release # build dist/plugin for ChatGPT Work
Layout: server/ (config, discovery, SSAP client, input socket, Wake-on-LAN, MCP server), test/ (node:test suite plus test/fixtures/fake-tv.mjs, a scripted SSAP TV), skills/ (assistant-facing usage skill), docs/superpowers/ (original design spec and implementation plan).
Quirks worth knowing: the reported model string is a constant (LG 65NANO80VNA — the TV this was built for), and user-facing error messages are in Hebrew, since that's the household it serves. PRs welcome.
License
MIT © Omer Lapidot
<div dir="rtl">
שלט לטלוויזיה — עברית
תוסף מקומי שהופך את ChatGPT (או Claude) לשלט רחוק לטלוויזיית LG webOS ברשת הביתית: הדלקה וכיבוי, ווליום, פתיחת אפליקציות כמו נטפליקס ויוטיוב, החלפת כניסות HDMI, ערוצים, הקלדת טקסט, הודעות על המסך וצילום של מה שמוצג על המסך.
הכול מקומי. התקשורת היא ישירות מהמחשב לטלוויזיה ברשת הביתית (websocket מקומי + Wake-on-LAN). שום דבר לא עובר בענן.
התקנה
המדריך המלא: docs/client-installation-he.md. בקצרה:
- מתקינים את התוסף שלט לטלוויזיה ביישום ChatGPT Work למחשב.
- מפעילים מחדש את ChatGPT, פותחים משימה וכותבים: "מה מצב הטלוויזיה?".
- בפקודת השליטה הראשונה יופיע על מסך הטלוויזיה חלון אישור — מאשרים פעם אחת עם השלט הפיזי, וזהו.
- כדי שההדלקה מרחוק תעבוד, מפעילים בטלוויזיה את ההגדרה Turn on via Wi-Fi.
</div>
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.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
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.