MCP AI Bridge
Control your Homey Pro with any AI assistant. Claude, ChatGPT, Gemini, Copilot or Cursor.
Just ask in plain language.
Talk to your Homey like you talk to a person. “Turn off the lights.” “Is the door still open?” “Set the heating to 21.” “Build me a flow for the porch light at sunset.” MCP AI Bridge lets your favorite AI assistant control your Homey Pro directly, in any language you speak. It runs locally on your Homey, works with every major assistant through the open MCP standard, comes with a full interactive home dashboard, and you stay in full control.
Links
App Store: MCP AI Bridge | Homey
Test version: MCP AI Bridge | Homey
Support and feedback: right here in this topic
What It Is
Your Homey already runs your home. Now you can just talk to it.
MCP AI Bridge turns your Homey Pro into something any modern AI assistant can talk to directly. No middleman cloud, no separate integration per assistant. The app runs a local Model Context Protocol (MCP) server on your Homey, and any MCP-compatible AI connects straight to it over your own network.
MCP is the open standard Anthropic, OpenAI and Google are building their assistants around. Because the bridge speaks that standard, the same Homey endpoint works in Claude, ChatGPT, Gemini, Cursor, Windsurf and more. No lock-in.
Why This MCP App
Works with every assistant, not one. One open-standard endpoint for Claude, ChatGPT, Gemini, Cursor, Windsurf and any MCP client.
Covers the whole Homey. Over a hundred actions across devices, zones, flows, advanced flows, logic, energy, insights, alarms, notifications, apps and system.
A full interactive dashboard, built in. Auto-generated or AI-built, with real tile control, in any browser on your network.
You build your own AI commands in the flow editor and send live answers back.
Local first. The AI connects to your Homey’s own address. Nothing routed through an extra cloud.
Real security. Scoped keys, an access switch, a tool blocklist, rate limits and a security log with source IPs.
Just Ask
Lights and climate: “Turn off all the lights downstairs.” · “Dim the living room to 30 percent.” · “Set the bedroom to 19 tonight.”
Check on things: “Is the garage door still open?” · “Which devices have a low battery?” · “Anything offline right now?”
Energy and history: “What used the most power this week?” · “Average power use over the last 7 days?”
Automations: “Make a flow that turns on the porch light at sunset.” · “Watch the front door and tell me when it opens.”
Housekeeping: “Did anything fail in the last hour?” · “Restart the Hue app.” · “Are any of my flows broken?”
You talk normally. The assistant works out which devices and capabilities you mean and does it.
What It Can Reach
Devices and every capability they have- Zones and rooms
Basic flows and Advanced Flows, including creating and editing them
Logic variables
Alarms and the alarm clock- Energy, live and historical
Insights, with one-number summaries
Notifications
Installed apps (enable, disable, restart, settings)
Users and presence
System info, memory, storage and reboot
Over a hundred actions in total, so the assistant can both understand the state of your home and act on it.
The Home Dashboard
Beyond chat, the app serves a complete, interactive dashboard in any browser on your network, at http://YOUR-HOMEY-IP:52199/dashboard.
Two ways to get one
One-tap auto-generate (no AI needed). The app reads your devices and builds a clean, tabbed dashboard grouped by what you have.- Build it with any AI. A structured config mode means Claude, ChatGPT, Gemini etc. all produce the same clean result. Just ask, or copy the ready-made prompt from the settings.
What you get
Real control from tiles: tap to toggle a light, long-press to dim, set a thermostat (temperature + mode), play or pause a speaker, lock the door, or run a one-tap scene like “all off”. Everything stays in sync with the live state.
Organised tabs: a live Overview (who is home, lights on, what’s playing, what needs attention, weather, live energy flow), Lights per room, Climate, Energy (electricity, water and gas for today/yesterday/month/year plus savings), Media, Security, a dedicated Vehicles tab for EVs (battery, range, lock, charge status, odometer), and a fallback Devices tab so nothing gets left off.
Self-updating: add or remove a device in a category and an open dashboard refreshes on its own.
Camera friendly: a plain <img>shows live snapshots and a<video>plays a stream (for formats the browser supports natively) on a custom HTML dashboard.
Multiple dashboards: keep several side by side (each at /dashboard/<name>) and choose which one opens by default.- Safe by design: controls use a separate dashboard token, are limited to the devices the dashboard shows, and can never delete or reboot anything.
How It Works
Under the hood it is a standard MCP server speaking JSON-RPC over local HTTP, using the StreamableHTTP transport from the MCP spec. Prefer to script against it? There is a REST shortcut at POST /tools/{name} and a full OpenAPI 3.1 spec at GET /openapi.json, so it drops right into n8n, Make, or your own code. Everything stays on your network.
Connect a Local Assistant
Desktop and CLI assistants run on your own machine, so they reach your Homey directly. Every one connects to the same address:
http://YOUR-HOMEY-IP:52199/mcp
Open MCP AI Bridge → Settings → Connection for the exact URL and a ready-to-paste snippet. The examples use <homey-ip>.
API key. If you set one in Settings → Security, add an
Authorization: Bearer YOUR_KEYheader (shown per client). Without a key the server is open on your local network only.
Claude Desktop — Settings → Developer → Edit Config, then restart:
{ "mcpServers": { "homey": { "type": "http", "url": "http://<homey-ip>:52199/mcp" } } }
Claude Code (CLI)
claude mcp add --transport http homey http://<homey-ip>:52199/mcp
Cursor — Settings → MCP → Add new server (or ~/.cursor/mcp.json):
{ "mcpServers": { "homey": { "url": "http://<homey-ip>:52199/mcp" } } }
Windsurf — ~/.codeium/windsurf/mcp_config.json:
{ "mcpServers": { "homey": { "serverUrl": "http://<homey-ip>:52199/mcp" } } }
GitHub Copilot (VS Code) — .vscode/mcp.json (note: servers, not mcpServers), then Agent mode:
{ "servers": { "homey": { "type": "http", "url": "http://<homey-ip>:52199/mcp" } } }
Gemini CLI — ~/.gemini/settings.json:
{ "mcpServers": { "homey": { "httpUrl": "http://<homey-ip>:52199/mcp" } } }
Antigravity CLI (replaces gemini-cli) — ~/.gemini/antigravity-cli/mcp_config.json, uses serverUrl:
{ "mcpServers": { "homey": { "serverUrl": "http://<homey-ip>:52199/mcp" } } }
Fallback: mcp-remote (for stdio-only clients, needs Node.js):
{ "mcpServers": { "homey": { "command": "npx", "args": ["-y", "mcp-remote", "http://<homey-ip>:52199/mcp", "--allow-http"] } } }
Quick check: ask “List my zones.” If that works, you are good to go.
Remote Access for Cloud & Mobile Assistants (v2.6, opt-in)
Cloud assistants (the ChatGPT and Claude web/phone apps) run on the provider’s servers, so they cannot see a private 192.168.x.x address. As of 2.6.0 the bridge is a full remote MCP server with standard OAuth 2.1, so you can use it from your phone, no desktop needed.
Step by step
- Set an API key in Settings → Security (you use it to approve the connection).
- Make your Homey reachable over HTTPS, e.g. a free Cloudflare Tunnel to port
52199. Never expose a bare port-forward without HTTPS. - In the app, open the new Settings → Remote tab: enter your public URL, turn remote access on, and save.
- In the Claude or ChatGPT app, add a custom connector with that URL +
/mcp. The app runs the OAuth login itself. - You get an approval screen: enter your API key once to confirm. Done, your Homey now works inside the AI app.
Secure by default: remote access is off until you enable it. With the switch off, the app stays local-network only, exactly as before, with no OAuth endpoints exposed. When on, it uses PKCE, single-use authorization codes and tokens that expire automatically.
Build Your Own AI Commands with Flow Cards
The app ships flow cards so your AI and your flows work together both ways.
Give the AI its own command. Trigger “An AI agent calls a custom tool”, run any logic, and reply with “Return a response to the AI agent” using live device values.
See what the AI does. Trigger “The AI controls a device” with name, capability and value as tokens.
Catch problems early. Triggers for a flow becomes broken, an AI tool call fails, a security event occurs, plus a “there is a broken flow” condition.
Give the AI memory. Action “Remember a value for the AI” and condition “A remembered value is”.
You design your own AI commands in the flow editor.
Real World Examples
- Answer with live data. Trigger An AI agent calls a custom tool → Return a response with the current power usage. Now “how much power am I using right now” gets the real number from your flow.
Let the AI know nobody is home. Your presence flow writes nobody_home = true; the assistant checks that context later.
Get told when a flow breaks. Trigger A flow becomes broken → send yourself a notification with the flow name.
You Stay in Control
Access mode switch: full, write only, read only, or off in one click.- Scoped API keys: each assistant its own key with read, write or full rights.
Tool blocklist: disable risky tools entirely, even for a full key.
IP whitelist and rate limits, with a tighter limit for destructive actions.
Security log: every denied or unauthorized attempt, with its source IP.
Active connections view. Argument values are never logged.
By default, with nothing configured, it stays open on your trusted local network. The controls are there when you want them.
See What Is Happening
The settings page is organized into tabs: Status, Connection, Commands, Dashboard, Security, Remote, Memory, Schedules, Events, Diagnostics, Bug report and About. Browse what the AI remembers and schedules, which devices it watches, recent activity, broken flows and the security log.
Works With
Claude (Desktop, Code, web, mobile), ChatGPT, Google Gemini, GitHub Copilot, Cursor, Windsurf, and any other MCP client. Text-to-speech and LED ring tools work on Homey Pro (Early 2019) and older models.
Get It
- Install: MCP AI Bridge | Homey
- Or the test build: MCP AI Bridge | Homey
- Open the app settings and run the connect wizard.
- Paste the config into your assistant and start talking to your home.
Questions, ideas and bug reports are very welcome in this topic. 
Changelog
v2.6.0 Remote access for cloud AI apps (opt-in). New "Remote" tab: OAuth 2.1
so Claude/ChatGPT apps connect over the internet. Off by default; when
on, connector registers, you approve once with your API key, tokens
expire automatically. PKCE, single-use codes, hashed tokens. With the
switch off the app stays LAN-only, exactly as before.
v2.5.4 Dashboards can play external video/audio (media-src): a plain <video>/
<audio> to an http/https stream works directly. Scripts/connections
stay restricted to the app.
v2.5.3 Dashboards can show external images (img-src): a plain <img> from any
http/https source works directly (camera snapshots, weather graphics).
v2.5.2 Electric cars get their own Vehicles tab (battery, range, lock, charge,
odometer); home battery stays on Energy. Every device type gets a home
(fans->Climate, TV/amp->Media, garage/camera->Security) + a Devices
fallback. Auto dashboards self-refresh on device add/remove. Charge/
lock/contact shown as readable text; several frozen rows now live.
v2.5.1 Dashboard hardening: page token limited to shown devices + honours the
access mode; long-press dimmer fixes; no more "NaN"; flow-create
validates + requires a trigger; no double-run on device error; logic
type fix; 0-100 volume; alarm repeat-days apply; webhook SSRF block;
settings survive restart; malformed requests handled.
v2.5.0 Full interactive dashboard: one-tap auto-generate or AI-built, live
control from tiles, tabs for lights/climate/energy/media/security,
multiple dashboards, separate dashboard token.
v2.4.0 Moods & scenes, weather + forecast, zone activity, realtime events,
AI-generated dashboard at /dashboard, modern MCP (structured output,
resource links, subscriptions, paging).
v2.3.0 ~93% smaller full-home overview, compact-by-default reads, control by
name, short read cache, MCP resources & prompts, expanded Connect tab.
v2.2.x Connect-tab guides (Gemini/Antigravity, Copilot, ChatGPT, Claude web),
auto Authorization header, Advanced Flow create/edit fixes, token
session fallback for privileged actions, my.homey.app settings fixes,
constant-time key compare, Commands tab.
v2.1.x New flow cards (AI controls a device, flow broken, tool fails, security
event, AI memory), Bug report tab + tool, stricter-client compatibility,
deep security & reliability audit.
v2.0.0 AI memory, device health/find/batch, insight summaries, scheduling,
events & webhooks, templates + flow validator, access modes, scoped
keys, blocklist, rate limits, security log, rebuilt NL/EN settings.
v1.x Renamed to MCP AI Bridge, custom-tool flow cards + response callback,
API key auth, REST + OpenAPI, Gemini support. v1.0.0: first release,
control Homey Pro from any AI agent over MCP.
