Skip to main content

n8n Integration - NotebookLM MCP HTTP Server

Complete guide for integrating NotebookLM into your n8n workflows

🎯 Overview​

This guide shows you how to query NotebookLM from n8n to create powerful automation workflows.

What you'll learn:

  • HTTP Request node configuration
  • Complete workflow examples
  • Error handling
  • Best practices

📋 Prerequisites​

  1. ✅ NotebookLM MCP HTTP server running on your Windows PC
  2. ✅ n8n installed (Docker, npm, or cloud)
  3. ✅ Network: Windows PC and n8n server can communicate

🔧 Basic Configuration​

Step 1: Find the Server IP​

On the Windows PC where the server is running:

ipconfig

Note the IPv4 (e.g., 192.168.1.52)

Step 2: Test the Connection​

From the n8n server:

curl http://192.168.1.52:3000/health

If it works, proceed to step 3. Otherwise, see Troubleshooting.

Step 3: Create the HTTP Request Node in n8n​

  1. Add a node "HTTP Request"

  2. Configure:

    • Method: POST
    • URL: http://192.168.1.52:3000/ask
    • Authentication: None
    • Send Body: Yes
    • Body Content Type: JSON
    • Specify Body: Using Fields Below

  3. Body Parameters (Add Field):

    • question → Type: String → Value: {{ $json.query }}
    • notebook_id → Type: String → Value: parents-numerique

📚 Example Workflows​

Workflow 1: Webhook → NotebookLM → Response​

Use case: REST API that queries NotebookLM

┌──────────────┐     ┌────────────────┐     ┌──────────────┐
│ Webhook      │────▶│ HTTP Request   │────▶│ Respond      │
│ (Trigger)    │     │ (NotebookLM)   │     │ to Webhook   │
└──────────────┘     └────────────────┘     └──────────────┘

Webhook Configuration:

  • Method: POST
  • Path: /ask-notebooklm

HTTP Request Configuration:

  • URL: http://192.168.1.52:3000/ask
  • Body: 
    {
      "question": "{{ $json.body.question }}",
      "notebook_id": "{{ $json.body.notebook_id }}"
    }

Respond Configuration:

  • Response Code: 200
  • Response Body: 
    {
      "answer": "{{ $json.data.answer }}",
      "session_id": "{{ $json.data.session_id }}"
    }

Test:

curl -X POST http://n8n-server:5678/webhook/ask-notebooklm \
  -H "Content-Type: application/json" \
  -d '{"question":"Test","notebook_id":"parents-numerique"}'

Workflow 2: Schedule → NotebookLM → Email​

Use case: Automatic daily report

┌──────────────┐     ┌────────────────┐     ┌──────────────┐
│ Schedule     │────▶│ HTTP Request   │────▶│ Send Email   │
│ (Cron)       │     │ (NotebookLM)   │     │              │
└──────────────┘     └────────────────┘     └──────────────┘

Schedule Configuration:

  • Trigger Times: Cron Expression
  • Expression: 0 9 * * 1-5 (9am Monday to Friday)

HTTP Request Configuration:

  • URL: http://192.168.1.52:3000/ask
  • Body: 
    {
      "question": "Quels sont les principaux conseils pour les parents?",
      "notebook_id": "parents-numerique"
    }

Email Configuration:

  • To: team@example.com
  • Subject: Rapport Parents & Numérique - {{ $now.format('DD/MM/YYYY') }}
  • Text: {{ $json.data.answer }}

Workflow 3: Form Submit → NotebookLM → Slack​

Use case: Slack bot that answers questions

┌──────────────┐     ┌────────────────┐     ┌──────────────┐
│ Slack        │────▶│ HTTP Request   │────▶│ Slack        │
│ (Trigger)    │     │ (NotebookLM)   │     │ (Send Msg)   │
└──────────────┘     └────────────────┘     └──────────────┘

Slack Trigger Configuration:

  • Event: App Mention
  • Workspace: Your workspace

HTTP Request Configuration:

  • URL: http://192.168.1.52:3000/ask
  • Body: 
    {
      "question": "{{ $json.event.text }}",
      "notebook_id": "parents-numerique"
    }

Slack Send Configuration:

  • Channel: {{ $json.event.channel }}
  • Text: {{ $('HTTP Request').item.json.data.answer }}

🔐 Security in n8n​

On the Windows PC, only allow the n8n IP:

New-NetFirewallRule `
  -DisplayName "NotebookLM (n8n only)" `
  -Direction Inbound `
  -LocalPort 3000 `
  -Protocol TCP `
  -Action Allow `
  -RemoteAddress 192.168.1.100  # n8n IP

Option 2: API Key​

If you have enabled the API key (see 02-CONFIGURATION.md):

In the HTTP Request node, add a Header:

  • Name: Authorization
  • Value: Bearer votre-api-key-secrete

🎨 Best Practices​

1. Error Handling​

Add an "Error Trigger" node after HTTP Request:

HTTP Request ──✓──▶ Success Path
             └─✗──▶ Error Trigger ──▶ Send Error Notification

Error Trigger Configuration:

  • Error Message: {{ $json.error }}
  • Action: Send notification, log, retry, etc.

2. Timeout​

NotebookLM can take 30-60 seconds. Configure the timeout:

HTTP Request → Settings:

  • Timeout: 120000 (120 seconds)

3. Rate Limiting​

Free NotebookLM = 50 requests/day. Add a counter:

┌────────────┐     ┌──────────────┐     ┌────────────┐
│ Check      │────▶│ HTTP Request │────▶│ Increment  │
│ Counter    │     │              │     │ Counter    │
└────────────┘     └──────────────┘     └────────────┘

4. Reusing Sessions​

For follow-up questions, pass the session_id:

First question:

{
  "question": "Quels conseils pour gérer le temps d'écran?"
}

Follow-up questions:

{
  "question": "Donne-moi un exemple",
  "session_id": "{{ $('HTTP Request 1').item.json.data.session_id }}"
}

📊 Useful Variables in n8n​

VariableDescriptionExample
{{ $json.data.answer }}NotebookLM responseResponse text
{{ $json.data.session_id }}Session ID"9a580eee"
{{ $json.data.session_info.message_count }}Message count3
{{ $json.success }}Success/failuretrue/false
{{ $json.error }}Error message"Timeout..."

🐛 Troubleshooting​

Problem 1: "ECONNREFUSED"​

Cause: n8n cannot reach the server

Solutions:

  1. Verify the server is running: curl http://192.168.1.52:3000/health
  2. Check the IP (not localhost from remote n8n!)
  3. Check Windows firewall

Problem 2: "Timeout"​

Cause: NotebookLM response too long

Solutions:

  1. Increase timeout HTTP Request → Settings → Timeout: 120000
  2. Verify NotebookLM responds: test with curl

Problem 3: "403 Forbidden"​

Cause: Invalid or missing API key

Solutions:

  1. Check Authorization header
  2. Check the API key value

Problem 4: Empty Responses​

Cause: Invalid notebook_id or notebook not configured

Solutions:

  1. List notebooks: GET /notebooks
  2. Use the correct ID or complete URL

📖 Advanced Examples​

Multi-Notebook Workflow​

Query multiple notebooks based on topic:

Webhook ──▶ Switch ──┬──▶ HTTP (Notebook Parents)
                     ├──▶ HTTP (Notebook Génétique)
                     └──▶ HTTP (Notebook Shakespeare)

Switch Configuration:

  • Mode: Rules
  • Rules:
    • If {{ $json.topic }} equals "parents-numerique" → Output 1
    • If {{ $json.topic }} equals "genetique-sante" → Output 2
    • Otherwise → Output 3

Workflow with Validation​

Validate the question before querying NotebookLM:

Webhook ──▶ Validate ──✓──▶ HTTP Request ──▶ Response
                      └─✗──▶ Error Response

Validate Configuration (Function node):

if (!items[0].json.question || items[0].json.question.length < 3) {
  throw new Error('Question trop courte');
}
return items;

🎓 Resources​

n8n integration complete! ✅