Development Plan: /slide Command - NotebookLM MCP Integration

Overview

Build a /slide command for KnowledgeFactory that sends Obsidian notes to NotebookLM via MCP server, generates multimedia assets (infographics, slides, audio), and returns shareable links.

Goal: Transform any vault note into visual/audio content via NotebookLM MCP integration, enabling one-command content generation.

MCP Integration Approach

Instead of browser automation, we use the Model Context Protocol (MCP) to directly control NotebookLM programmatically. This is cleaner, more reliable, and supports all NotebookLM features.

Why MCP over Browser Automation

The 5-Level Workflow

Based on the AntiGravity + NotebookLM tutorial:

MCP Server Options

Option A: Python Version (Recommended for features)

Repository: jacob-bd/notebooklm-mcp

Best for: Advanced users who want full control, including Audio Overviews, Slide Decks, and Deep Research mode.

Installation:

uv tool install notebooklm-mcp-server

Features:

• Create and manage notebooks
• Add sources (text, URLs, files)
• Generate Audio Overviews
• Generate Slide Decks
• Deep Research mode
• Full CRUD operations

Option B: Node.js Version (Recommended for speed)

Repository: PleasePrompto/notebooklm-mcp

Best for: Quick setup and using NotebookLM primarily for retrieving answers with citations to prevent AI hallucinations.

Installation:

npx notebooklm-mcp@latest

Features:

• Fast notebook operations
• Citation-backed responses
• Lightweight setup
• Quick query/response cycle

Implementation Plan

Phase 1: MCP Server Setup

1.1 Install NotebookLM MCP Server

# Option A: Python (full features)
uv tool install notebooklm-mcp-server

# Option B: Node.js (quick setup)
npx notebooklm-mcp@latest

1.2 Configure Claude Code MCP Settings

Add to .claude/settings.json or global MCP config:

{
  "mcpServers": {
    "notebooklm": {
      "command": "notebooklm-mcp-server",
      "args": []
    }
  }
}

1.3 Authentication Setup

• Google account authentication (one-time browser auth)
• Credentials cached for subsequent API calls

Phase 2: /slide Command Implementation

2.1 Slash Command Definition

Location: .claude/commands/slide.md

# /slide - Generate Multimedia from Notes via NotebookLM MCP

Generate infographics, slides, audio overviews, and study guides from vault notes using NotebookLM MCP integration.

## Usage

/slide <filename.md> [--type <asset-type>]

## Arguments

- `filename.md` - The note to convert (relative to vault root)
- `--type` - Asset type: slide-deck (default), audio, study-guide, research

## Asset Types

| Type | Output | Use Case |
|------|--------|----------|
| slide-deck | Presentation slides | Visual presentations |
| audio | Audio Overview (MP3) | Podcast-style listening |
| study-guide | Structured outline | Learning materials |
| research | Deep Research report | Comprehensive analysis |

## Workflow

1. Read vault file via MCP Obsidian tools
2. Create NotebookLM notebook via MCP
3. Add note content as source
4. Trigger asset generation (slides/audio/etc)
5. Return shareable link

## Examples

# Generate slide deck (default)
/slide my-research-note.md

# Generate audio overview
/slide quarterly-report.md --type audio

# Generate deep research
/slide findings.md --type research

2.2 MCP Tool Flow

STEP 1: Read vault file
├─ Use MCP Obsidian tools: get_vault_file("my-note.md")
└─ Extract content and metadata

STEP 2: Create NotebookLM Notebook
├─ Call: notebooklm_create_notebook(title)
└─ Returns: notebook_id

STEP 3: Add Source Content
├─ Call: notebooklm_add_source(notebook_id, content)
└─ Wait for processing

STEP 4: Generate Asset
├─ Call: notebooklm_generate_slides(notebook_id)
│   OR: notebooklm_generate_audio(notebook_id)
│   OR: notebooklm_deep_research(notebook_id, query)
└─ Returns: asset_url

STEP 5: Return Result
├─ Output shareable link
└─ Success message with notebook URL

Phase 3: Integration with KnowledgeFactory

3.1 Update obsidian-vault-manager Skill

Add /slide command to SKILL.md:

| /slide | Generate slides/audio from notes | /slide note.md --type slide-deck |

3.2 Output Queue Integration

File: .github/data/queue/out-sld-{id}.json

{
  "type": "output",
  "command": "slide",
  "sourceNote": "2026-01-20-my-note.md",
  "assetType": "slide-deck",
  "cleanup": true,
  "cleanupDays": 30,
  "addedAt": "2026-01-20T10:00:00Z"
}

Architecture Diagram

User's Obsidian Vault
        │
        │ /slide my-note.md --type slide-deck
        ▼
Claude Code (/slide command)
        │
        ├─ Read vault file (MCP Obsidian Tools)
        │
        ├─ Call NotebookLM MCP Server
        │
        ▼
NotebookLM MCP Server
        │
        ├─ Create notebook
        │
        ├─ Add note content as source
        │
        ├─ Generate slide deck / audio / research
        │
        └─ Return shareable link
        │
        ▼
Return to user
        │
        ├─ Asset link: https://notebooklm.google.com/...
        │
        ├─ Notebook link: https://notebooklm.google.com/...
        │
        └─ Success message

Development Tasks

MVP (Minimum Viable Product)

• Install and configure NotebookLM MCP server (Python version)
• Test MCP connection with simple notebook creation
• Create .claude/commands/slide.md command definition
• Implement slide-deck generation workflow
• Test end-to-end with sample note
• Document in obsidian-vault-manager SKILL.md

Enhanced Version (v1.1)

• Support all asset types:
  • slide-deck
  • audio (Audio Overview)
  • study-guide
  • research (Deep Research)
• Add progress/status messages
• Error handling for generation failures
• Return both asset link AND notebook URL

Future Enhancements (v2.0+)

• Bulk file ingestion (Level 5)
• Interactive dashboard generation (Level 3)
• Auto-cleanup old notebooks
• Integrate with DoubleCopy output queue
• Mobile trigger via DoubleCopy Mobile app
• Citation extraction for fact-checking

Testing Checklist

• MCP server connects successfully
• Notebook creation works
• Source content uploads correctly
• Slide deck generation completes
• Audio overview generation completes
• Deep research mode works
• Shareable links are accessible
• Error handling works for failures
• Full process completes in reasonable time

Resources

Tutorial Video

• Build a NotebookLM App with AntiGravity - Step-by-step setup guide

MCP Server Repositories

• jacob-bd/notebooklm-mcp - Python version (full features)
• PleasePrompto/notebooklm-mcp - Node.js version (speed)

Related Documentation

• [[KnowledgeFactory Enterprise]] - Parent system documentation
• NotebookLM - Google NotebookLM
• Model Context Protocol - MCP specification
• obsidian-vault-manager skill - Integration target

Created: 2026-01-20 Updated: 2026-01-25 Status: Planning → Ready for MCP Implementation Next Action: Install NotebookLM MCP server and test basic notebook creation