Documentation › Web Chat

Web Chat

The Web Chat is a built-in AI chat interface embedded in the BunkerM dashboard. It connects you to BunkerAI so you can ask questions, give commands, and control your broker with plain language - no Slack workspace or Telegram setup required.

How to Access

  • Click Ask BunkerAI in the left sidebar.
  • Or navigate to Tools > AI Chat.

The chat opens with your conversation history above and a message input at the bottom.

Requires BunkerAI Cloud. Connect your API key in Settings > Integrations. The Community plan includes a starter credit allocation.

Control Your Entire IoT Setup With Plain English

BunkerAI gives you full, natural-language control over your MQTT infrastructure. You do not need to know topic structures, payload formats, QoS levels, or ACL syntax. Just describe what you want — BunkerAI figures out the rest.

Real example You say: "Turn off my room light"
BunkerAI looks up which device controls that light from your topic annotations and shared memory, determines the correct off payload (e.g. 0 or OFF), and publishes it to the right topic — all without you knowing the topic path or payload format.

What BunkerAI Can Do

BunkerAI has full access to all BunkerM APIs and can take real actions on your running Mosquitto broker. The scope is intentionally broad — if it can be done in BunkerM, it can be done through plain English conversation.

Device and topic control

Turn off the kitchen lights.
Set the thermostat to 21 degrees.
What is the current temperature in the living room?
Publish a restart command to the gateway.
What is the last message on home/sensor/door?

BunkerAI uses your topic annotations and conversation history to map plain-language device names to the correct MQTT topics and payloads. The more you annotate your topics in Settings → Annotations, the smarter it gets.

Client and access control

Create 10 sensor clients with random credentials.
Disable client broken-device-03 immediately.
Give gateway-01 publish access to all sensor topics.
Create a role called read-only that can only subscribe.
Show me all clients that haven't connected in the last 7 days.
Block all clients in the legacy-devices group.

Broker monitoring and diagnostics

How many clients are connected right now?
What topics are most active?
Summarize anomalies from the last 24 hours.
Is there anything unusual happening on my broker?
Which clients are generating the most traffic?

Broker configuration

Set the maximum number of connections to 500.
Enable TLS on port 8883.
What is my current broker configuration?
Increase the message size limit to 10MB.

Automation agents

Alert me on Telegram when the temperature exceeds 35°C.
Create a heartbeat that publishes "online" to system/status every minute.
Pause all scheduler jobs until tomorrow morning.
Show me all active watchers and their current state.

Explanations and guidance

Why was client sensor-01 disconnected?
Explain what MQTT QoS 2 means in practice.
What is the # wildcard and when should I use it?
Why is my retained message not being received?

The AI Knows Your Full Broker State

BunkerAI is not just answering general questions — it has direct, live access to your broker's current state on every single message. This includes:

  • All active MQTT topics with their latest published payloads
  • Connected clients — who is online right now and their metadata
  • Full ACL configuration — every client, role, group, and their permissions
  • Broker statistics — message rates, uptime, subscriptions, byte throughput
  • Anomaly history — current alerts and past detection events
  • Topic annotations — your human-readable device labels and descriptions
  • Conversation memory — context from previous sessions and other channels (Telegram, Slack)

Actions are executed immediately and for real. When you say "create 10 clients", they appear in Mosquitto's dynamic security. When you say "disable that sensor", it is disabled instantly. When you ask "what is the temperature?", you get the live retained value — not a generic answer.

Tips for Better Results

  • Annotate your topics. Go to Settings → Annotations and label your topics with human-readable names. This is what lets you say "turn off the kitchen light" instead of knowing the exact topic path.
  • Use natural device names. Once annotated, you can refer to "the front door sensor", "the thermostat", or "all lights in the living room" directly.
  • Chain requests. "Create 5 clients, add them all to the sensors group, and give that group publish access to sensor/#" works as a single message.
  • Ask for context. "Why did you do that?" or "What just changed?" will give you a plain-English explanation of every action taken.
  • Continue across channels. A conversation started on Telegram or Slack carries over to Web Chat automatically — no need to re-explain your setup.

Clearing Conversation History

  1. Open the Web Chat.
  2. Click Clear History or New Conversation.
  3. Confirm the clear.

After clearing, the AI starts fresh without memory of previous conversations. Note that topic annotations are not affected - they persist separately.

← Slack Shared Memory →