Skip to main content

Node-RED Configuration

Node-RED flow for Claude AI integration.

Access Node-RED​

Flow Overview​

┌─────────────┐    ┌──────────────┐    ┌─────────────┐
│  MQTT In    │───►│ Detect       │───►│ Claude API  │
│  (receive)  │    │ @claude      │    │ Request     │
└─────────────┘    └──────────────┘    └──────┬──────┘
                                              │
┌─────────────┐    ┌──────────────┐    ┌──────▼──────┐
│  MQTT Out   │◄───│ Format       │◄───│ Process     │
│  (send)     │    │ Response     │    │ Response    │
└─────────────┘    └──────────────┘    └─────────────┘

Nodes Configuration​

1. MQTT In Node​

Receives messages from the mesh network.

Settings:

  • Server: 192.168.68.127:1883
  • Topic: meshtastic/2/json/LongFast/!69d01ebc
  • QoS: 0
  • Output: auto-detect

2. Detect @claude (Function Node)​

Filters messages containing "@claude".

var payload = msg.payload;

// Parse if string
if (typeof payload === 'string') {
    try {
        payload = JSON.parse(payload);
    } catch (e) {
        return null;
    }
}

// Check for text message with @claude
if (payload.payload && payload.payload.text) {
    var text = payload.payload.text;
    if (text.toLowerCase().includes('@claude')) {
        msg.originalSender = payload.sender;
        msg.question = text.replace(/@claude/gi, '').trim();
        return msg;
    }
}
return null;

3. Claude API Request (HTTP Request Node)​

Sends request to Claude API.

Settings:

  • Method: POST
  • URL: https://api.anthropic.com/v1/messages
  • Headers:
    • x-api-key: [API KEY]
    • anthropic-version: 2023-06-01
    • content-type: application/json

4. Prepare Request (Function Node)​

Formats the request body for Claude.

msg.payload = {
    "model": "claude-sonnet-4-20250514",
    "max_tokens": 150,
    "system": "You are a helpful assistant for a rural community. Give brief, clear answers. Maximum 2-3 sentences. Respond in the same language as the question.",
    "messages": [
        {
            "role": "user",
            "content": msg.question
        }
    ]
};

msg.headers = {
    "x-api-key": env.get("CLAUDE_API_KEY"),
    "anthropic-version": "2023-06-01",
    "content-type": "application/json"
};

return msg;

5. Process Response (Function Node)​

Extracts Claude's response.

if (msg.payload && msg.payload.content && msg.payload.content[0]) {
    msg.response = msg.payload.content[0].text;
    return msg;
}
return null;

6. Format for MQTT (Function Node)​

Prepares message for mesh network.

// Truncate if too long (LoRa limit ~230 bytes)
var response = msg.response;
if (response.length > 200) {
    response = response.substring(0, 197) + "...";
}

msg.payload = response;
msg.topic = "meshtastic/2/json/LongFast/!69d01ebc/sendtext";

return msg;

7. MQTT Out Node​

Sends response back to mesh.

Settings:

  • Server: 192.168.68.127:1883
  • Topic: (set by previous function node)
  • QoS: 0
  • Retain: false

Environment Variables​

Set in Node-RED settings:

  • CLAUDE_API_KEY: Your Anthropic API key

Import Flow​

To import the complete flow:

  1. Open Node-RED
  2. Menu → Import
  3. Paste the flow JSON
  4. Click Import
  5. Deploy

Debugging​

Add debug nodes after each step to see message flow:

  1. After MQTT In - see raw messages
  2. After Detect @claude - see filtered messages
  3. After Claude API - see responses
  4. After Format - see final output