Claude Code + NotebookLM:
Infinite Memory

Install, Configure & Master AI-Powered Long-Term Memory

Daniel Wright • NavAIgate • dw@navaigate.dev • navaigate.dev

TODAY'S JOURNEY

Agenda

1The Problem: Claude's Amnesia

2The Solution: NotebookLM as Second Brain

3Live Install: NotebookLM Skill

4Authentication & Setup

5Key Commands & Quick Reference

6Use Case: Enrichment — Deep Research

7Use Case: Content Multiplication

8Use Case: The AI Brain & WrapUp Skill

9Adding to Claude Co-work

10Q&A

01 // THE PROBLEM

Claude Has Amnesia

!Every new session starts from zero

!Reading files to restore context burns tokens & money

!Memory is temporal — context fades as conversations grow

!No persistent recall across sessions

02 // THE SOLUTION

NotebookLM as Your Second Brain

1. Persistent Memory

Decisions & context survive every session

2. Cost Reduction

Offload research to NotebookLM (free tier)

3. Content Multiplication

One source becomes podcast, video, infographic

03 // CAPABILITIES

What NotebookLM Brings to Claude

Persistent Memory

Decisions & rationale across sessions

Personal CRM

Contacts, meetings, follow-ups

Decision Journal

Record the "why" behind choices

Content Multiplication

Podcast, video, infographic from one source

Deep Research

40-70 sources synthesised into insights

Meeting Intelligence

Store and query meeting transcripts

04 // THE SKILL FILE

Inside NotebookLMSkill.md

Frontmatter — name, description, trigger phrases

Step 0: Setup — Python check, venv, pip install

Authentication — Custom Playwright script (built-in notebooklm login can't work in Claude Code)

Quick Reference — Full command table

Autonomy Rules — Auto-run vs. ask-first

Error Handling — Common issues & fixes

05 // LIVE INSTALL

Step by Step

1Drop NotebookLMSkill.md into Claude Code

2Say: "Execute this skill"

3Claude checks Python version (needs 3.10+)

4Creates venv at ~/.notebooklm-venv

5Installs notebooklm-py[browser] via pip

6Runs playwright install chromium (headless browser for auth)

7Symlinks CLI to ~/bin/notebooklm

8Verify: notebooklm --help

06 // AUTHENTICATION

Why a Custom Login Script?

The Problem

The built-in notebooklm login needs you to press Enter in the terminal after signing in. Claude Code's bash tool can't do interactive input, so the browser opens and closes instantly.

The Solution

The skill writes a custom Python script using Playwright (installed in step 6). It opens Chromium, waits for a file-based signal instead of keyboard input, then captures your cookies.

THE FLOW

1Claude runs the custom script — Chromium opens

2You sign into Google in that browser

3Navigate to notebooklm.google.com

4Tell Claude: "I'm signed in"

5Claude sends a file signal → cookies captured

6Verify: notebooklm auth check

If auth fails: delete ~/.notebooklm/browser_profile and retry

07 // COMMANDS

Essential Commands

Task	Command
List notebooks	`notebooklm list`
Create notebook	`notebooklm create "Title"`
Add URL source	`notebooklm source add "https://..."`
Add YouTube	`notebooklm source add "https://youtube.com/..."`
Chat with notebook	`notebooklm ask "question"`
Deep research	`notebooklm source add-research "query" --mode deep`
Generate podcast	`notebooklm generate audio "instructions"`
Generate infographic	`notebooklm generate infographic`
Download	`notebooklm download audio ./output.mp3`

08 // USE CASE 1

Enrichment — Deep Research

1Have a conversation in Claude about any topic

2"Create a notebook from this conversation with deep research"

3NotebookLM ingests 40-70 sources automatically

4Ask for distilled insights and action items

5One API call → massive depth → zero extra token cost

08 // LIVE DEMO

Enrichment in Action

"Create a brand new notebook based on everything
in this conversation. Add deep research on this topic.
Once done, give me 10 distilled insights with
3 core actions I need to do immediately."

"NotebookLM did all the heavy lifting — deep research across dozens of sources — and Claude just made one API call to get the results. Zero extra token cost."

Pro tip: If research times out, follow up with: "Make sure you imported the deep research results"

09 // USE CASE 2

Content Multiplication

One source → unlimited outputs, all free

Type	Output	Time
Podcast	.mp3	10-20 min
Video	.mp4	15-45 min
Slide deck	.pptx	5-10 min
Infographic	.png	2-5 min
Report	.md	2-5 min
Quiz	.html	5-15 min
Mind map	.json	Instant

09 // LIVE DEMO

Generate an Infographic

"Create an infographic in NotebookLM using
premium colours and fonts, based on our conversation.
Download it for me."

"Everything NotebookLM generates is free. Podcasts, videos, infographics — you're not burning Claude tokens for any of it."

Generation Cost

Output Types

API Call Needed

10 // USE CASE 3

The AI Brain

Persistent memory that grows with every session

1Reviews the entire conversation

2Saves key memories (decisions, learnings, preferences)

3Writes a structured session summary

4Pushes it to your AI Brain notebook in NotebookLM

10 // WRAPUP SKILL

Inside WrapUpSkill.md

Step 0: Auto-creates "AI Brain" notebook if none exists

Step 1: Reviews session — decisions, work, learnings

Step 2: Saves memories locally (feedback, project, user)

Step 3: Writes markdown session summary

Step 4: Pushes summary to NotebookLM as a source

Trigger Words

/wrapup • "wrap up" • "save this session" • "end of session"

/wrapup

11 // POWER MOVE

Auto-Consult Your AI Brain

Add this to your Claude project instructions:

"Whenever answering questions about strategy,
always consult the AI Brain notebook
in NotebookLM first."

One API Call

Instant semantic search of your entire history

Growing Brain

Every session adds more context — zero token cost to query

12 // CO-WORK

Adding to Claude Co-work

1In Claude Code: "Add NotebookLM as a Co-work skill please"

2Claude generates NotebookLMSkill-Cowork.md

3In Co-work: + → Skills → Manage Skills → Upload

4Upload the generated file

5Verify: "What was the last notebook I created?"

Why Different?

Co-work is sandboxed — can't read local files. Claude Code bakes auth cookies directly into the skill file.

Stripped to only essential cookies: ~1,400 tokens (vs ~3,100).

RECAP

Three Pillars

Enrichment

Deeper research, better decisions, less token spend

Content Multiplication

Free generation of podcasts, videos, infographics

Persistent Memory

AI Brain grows over time, semantic retrieval

"Your NotebookLM Brain is a semantic search engine over your entire Claude history — and it costs nothing to query."

Thank You & Q&A

Presenter

Daniel Wright

dw@navaigate.dev

Website

navaigate.dev

Resources

NotebookLM Skill file • WrapUp Skill file • Presenter guide

Video: "Claude Code + NotebookLM = Infinite Memory" — Jack Roberts

Claude Code + NotebookLM:Infinite Memory

Agenda

Claude Has Amnesia

NotebookLM as Your Second Brain

1. Persistent Memory

2. Cost Reduction

3. Content Multiplication

What NotebookLM Brings to Claude

Persistent Memory

Personal CRM

Decision Journal

Content Multiplication

Deep Research

Meeting Intelligence

Inside NotebookLMSkill.md

Step by Step

Why a Custom Login Script?

The Problem

The Solution

Essential Commands

Enrichment — Deep Research

Enrichment in Action

Content Multiplication

Generate an Infographic

The AI Brain

Inside WrapUpSkill.md

Trigger Words

Auto-Consult Your AI Brain

One API Call

Growing Brain

Adding to Claude Co-work

Why Different?

Three Pillars

Enrichment

Content Multiplication

Persistent Memory

Thank You & Q&A

Resources

Claude Code + NotebookLM:
Infinite Memory