Guide April 16, 2026

Using BunkerM AI Chat to Monitor and Control Your Smart Home

Every device in a smart home talks over MQTT, but reading that data usually means digging through dashboards, log files, or raw topic trees. BunkerM's AI chat connects directly to your live broker and lets you ask plain questions instead. What follows are real exchanges from a 10-device home setup, run through the web chat, and the same queries accessible from Telegram and Slack.

The Setup

The home setup used in these examples has 10 MQTT clients publishing to the broker: a home hub, kitchen lights, living room and bedroom sensors, a garage door, hallway motion detector, smart lock, garden irrigation system, attic leak detector, and an HVAC thermostat. All are standard MQTT clients publishing JSON payloads to topic paths like home/hvac/thermostat/state and home/entry/lock/events.

BunkerM sees all of this traffic in real time. The AI reads retained messages and live payloads on demand, so every answer reflects the current state of the broker, not a cached snapshot.

Conversation 1: Device Overview

Starting with the obvious question: what is the current state of everything?

Give me a summary of all connected devices in my smart home right now. What devices are online and what are they reporting?

Based on the live broker data, your smart home is currently active with several devices reporting status:

Climate & Environment

HVAC Thermostat: Running in cool mode (Setpoint: 20.3°C, Current: 20.9°C). Filter change recommended (69 days since last change).
Living Room Sensor: 20.8°C, 39.3% humidity, CO2 at 671 ppm.
Bedroom Sensor: 20.8°C, 47.4% humidity.
Attic: 27.7°C with 64.7% humidity.

Security & Access

Entry Lock: Currently locked. Last event: unlock by "family" via app.
Garage Door: Currently opening.
Attic Leak Detector: No leak detected.

Garden & Energy

Garden Irrigation: Inactive. Soil moisture at 39.1%.
Home Hub: Online. Uptime 403 hours, CPU at 9.5%, 10 devices registered.

In one message you get a full snapshot of the house. No dashboard tabs, no topic browsing. The AI reads the retained payloads directly from the broker and structures the answer.

Conversation 2: Energy Breakdown

Which devices are consuming the most power right now? Compare the HVAC and kitchen lights energy usage and tell me if anything looks abnormal.

Current Power Consumption

HVAC System: 1796W. Highest load in the home.
Kitchen Lights: 10.3W.

The HVAC is consuming approximately 174x more than the kitchen lights. A reading of 1796W is typical for a heat pump in active operation. COP is 2.8, within normal range. The 10.3W from kitchen LEDs is expected. Nothing looks abnormal.

This is the kind of comparison that normally requires pulling numbers from two different topic paths and doing the math manually. The AI does it inline and flags whether the numbers make sense given the operating context.

Conversation 3: Security Review

Review the smart lock and garage door activity. Has the front door been unlocked recently, and is the garage currently closed? Any suspicious access attempts?

Front Door: Currently locked. Last event was a lock action by "family" via auto method. There was 1 failed attempt today, but no ongoing suspicious activity.

Garage Door: Currently closing. Not fully closed yet. Previously closed for approximately 373 minutes before this movement.

The front door is secure. The garage is closing. The single failed lock attempt today does not indicate unauthorized access.

Conversation 4: Sensor Health Check

Are there any issues I should be aware of across all my sensors? Check the attic leak detector, HVAC filter status, and soil moisture in the garden.

⚠ HVAC Filter: 41 days since last change. change_recommended is currently false, but worth monitoring depending on your schedule.

✓ Attic Leak Detector: No leak. Humidity stable at 42.0%.

✓ Garden Soil Moisture: 41.7%. Healthy range. Irrigation correctly inactive.

Conversation 5: Climate Map and HVAC Performance

What's the current temperature and humidity across the house? Is the HVAC keeping up with the setpoint, and how does the living room compare to the bedroom?

Location	Temp	Humidity
Living Room	23.6°C	41.4%
Bedroom	18.9°C	45.4%
Attic	30.9°C	70.5%
Garage	12.5°C	49.5%

HVAC setpoint is 22.2°C. Current measured temp is 23.8°C. System is running with fan on high but still 1.6°C above target. The living room (23.6°C) and bedroom (18.9°C) have a 4.7°C gap, which points to uneven airflow or different thermal loads between rooms.

Conversation 6: Automation Suggestion

Based on the hallway motion sensor and living room occupancy data, can you suggest an automation rule to optimize the lighting schedule?

Living room is currently occupied at 90% confidence. Hallway motion is detected and lights are already ON. Based on this, I suggest a Conditional Occupancy Chain:

When home/hallway/motion/state.motion = true: turn hallway lights ON at 80% brightness.
When motion is absent for 2 minutes and living room occupancy drops below 50%: turn hallway lights OFF.
If living room remains occupied, keep hallway at 30% ambient instead of full off.

This reduces unnecessary full-brightness operation while keeping the hallway usable when someone is about to move through it.

Three Ways to Access It

Web Chat (in the BunkerM dashboard)

Open AI and then Chat in the sidebar. This gives you the full conversation interface with context preserved across the session. Useful when you are at your desk doing broker maintenance or reviewing device health. The AI can execute actions directly from here: publish messages, create or delete clients, set up watchers, and schedule jobs.

Connect a Telegram bot to BunkerM under Settings, then Integrations. Once configured, you can send any of the questions above directly to your bot from your phone. The AI replies with the same live broker data. Practical for quick checks while away from home: "Is the garage closed?" or "Did the leak detector trigger?" sent to a bot at 11 PM takes less than a second to answer.

# Example Telegram message to your BunkerM bot
Is the front door locked?
Is anything running in the garden right now?
What's the attic humidity?

Slack

For households or small teams that already use Slack, the Slack connector routes messages to the same AI. Connect your Slack workspace under Settings, then Integrations. From any channel where the bot is present, you can query the broker the same way. Useful for shared living situations or small rental properties where multiple people need visibility into the same infrastructure.

What the AI Actually Reads

On every query, BunkerM injects the following context into the AI request: the list of active MQTT clients, recent retained payloads from subscribed topics, broker statistics, topic annotations you have defined, and any watchers or scheduled jobs currently configured. This means the AI is not guessing. When it says the garage door is closing, it read home/garage/door/state at that moment.

It also understands your topic structure through annotations. If you label home/hvac/thermostat/state as "Main thermostat, JSON payload with temperature_c, setpoint_c, and mode fields", the AI can answer questions about it correctly without you explaining the schema every time.

Getting Started

Pull the latest BunkerM image and restart:

docker pull bunkeriot/bunkerm:latest
docker compose up -d

Connect a Telegram or Slack bot from Settings, then Integrations. For the web chat, it is available immediately after logging in. AI features require a BunkerAI Cloud connection or a local LLM configured under Settings, then Integrations, then Local LLM.

Full setup documentation is at bunkerai.dev/docs/bunkerai.

← All articles Read the docs →