If NOT found:

Install from pip:
```
pip install notebooklm-mcp-cli
```

Verify install:

PYTHONIOENCODING=utf-8 nlm --version 2>&1

If pip fails, tell user to run manually: ! pip install notebooklm-mcp-cli

0.2 Check if MCP server is registered in Claude Code

claude mcp list 2>&1 | grep -i notebooklm

If NOT found:

Find the binary path:

which notebooklm-mcp 2>/dev/null || where notebooklm-mcp 2>/dev/null

claude mcp add notebooklm-mcp -- "<full_path_to_notebooklm-mcp.exe>"

Tell the user: "MCP registered. You'll need to restart this Claude Code session for tools to appear."

0.3 Check if authenticated

PYTHONIOENCODING=utf-8 nlm login --check 2>&1

If auth is valid: proceed to the user's request.

If auth fails or no profile exists:

Ask the user: "Which Google account (email) do you want to use for NotebookLM?"
Wait for their answer. Store the email in a variable, e.g. user@example.com
Derive a profile name from the email (e.g. user@company.com → company)
Tell the user to run this in their terminal (it opens a browser — Claude Code can't do interactive browser auth):
```
! $env:PYTHONIOENCODING="utf-8"; nlm login --profile <profile_name>
```
Tell them: "Sign in with <their_email> in the browser that opens, then come back here and say 'done'."

When user confirms, verify:

PYTHONIOENCODING=utf-8 nlm login --check --profile <profile_name> 2>&1

Set as default:

PYTHONIOENCODING=utf-8 nlm login switch <profile_name> 2>&1

Confirm: "Connected as . You're ready to go."

0.4 Bootstrap complete

Once all 3 checks pass, proceed to the user's actual request.

When to Use

User wants to create a new NotebookLM notebook on a topic
User wants to add sources (URLs, text, Google Drive files) to a notebook
User wants to query/chat with notebook contents
User wants to run web research and import discovered sources
User wants to generate audio overviews (podcasts) from notebooks
User wants to list, describe, or manage existing notebooks
User says "notebook", "NotebookLM", "research notebook", "podcast from sources"

Environment Note (Windows)

All nlm commands MUST be prefixed with PYTHONIOENCODING=utf-8 to avoid Unicode rendering crashes on Windows cp1252 terminals.

Available Commands Reference

Notebook Management

nlm notebook list                          # List all notebooks (JSON)
nlm notebook create "Title"                # Create new notebook
nlm notebook get <id>                      # Get notebook details
nlm notebook describe <id>                 # AI-generated summary + suggested topics
nlm notebook rename <id> "New Title"       # Rename notebook
nlm notebook delete <id>                   # Delete permanently
nlm notebook query <id> "question"         # Chat with notebook sources

Source Management

nlm source list <notebook_id>              # List sources in notebook
nlm source add <notebook_id> --url "URL"   # Add URL source
nlm source add <notebook_id> --text "text" # Add text source
nlm source add <notebook_id> --drive "ID"  # Add Google Drive file
nlm source get <notebook_id> <source_id>   # Get source details
nlm source describe <notebook_id> <source_id>  # AI summary of source
nlm source content <notebook_id> <source_id>   # Raw source content
nlm source delete <notebook_id> <source_id>    # Delete source
nlm source stale <notebook_id>             # List Drive sources needing sync
nlm source sync <notebook_id>              # Sync Drive sources

Research (Web Discovery)

nlm research start <notebook_id> "query"   # Start web research task
nlm research status <notebook_id> <task_id> # Check research progress
nlm research import <notebook_id> <task_id> # Import discovered sources

Audio & Studio

nlm audio create <notebook_id>             # Generate podcast from sources
nlm studio status <notebook_id>            # List studio artifacts + status
nlm download <notebook_id> <artifact_id>   # Download audio/video/etc.

Cross-Notebook & Batch

nlm cross query "question"                 # Query across all notebooks
nlm batch <command> --all                  # Run command across notebooks
nlm tag add <notebook_id> "tag"            # Tag a notebook
nlm tag list                               # List all tags

Pipelines

nlm pipeline list                          # List available pipelines
nlm pipeline run <notebook_id> <pipeline>  # Execute pipeline on notebook
nlm pipeline create <yaml_file>            # Create custom pipeline

Utilities

nlm doctor                                 # Diagnose installation issues
nlm login --check                          # Verify auth status
nlm alias set <name> <id>                  # Create friendly alias for notebook ID

Workflow: Deep Research on a Topic

When the user asks to research a topic end-to-end:

Phase 1: Setup

Create a new notebook with a descriptive title
Save the notebook ID for subsequent commands

Phase 2: Seed Sources

If the user provides URLs, add them as sources
If the user provides text/documents, add as text sources
If Drive files are relevant, add via Drive ID

Phase 3: Web Research

Run nlm research start <id> "research query" to discover sources
Poll nlm research status until complete
Import the best discovered sources with nlm research import

Phase 4: Synthesize

Use nlm notebook query <id> "question" to extract insights
Run multiple queries to build a comprehensive picture
Summarize findings for the user

Phase 5: Outputs (Optional)

Generate audio podcast: nlm audio create <id>
Export to Google Docs: nlm export <id>
Create quiz/flashcards: nlm quiz <id> / nlm flashcards <id>

Workflow: Quick Add & Query

For quick knowledge base interactions:

nlm notebook list — find the target notebook
nlm source add <id> --url "..." — add the source
Wait a moment for processing
nlm notebook query <id> "question" — ask questions

Output Format

Always present results clearly:

When listing notebooks, show a clean table with title, source count, and last updated
When querying, present the NotebookLM response with any citations
When adding sources, confirm success and show the source ID
When research completes, summarize what was found before asking to import

Error Handling

If nlm returns auth errors → re-run Step 0.3 auth flow
If a notebook ID is not found → run nlm notebook list to show available notebooks
If research times out → check status again, it may still be processing
If Unicode errors appear → ensure PYTHONIOENCODING=utf-8 prefix is used

notebook-research

NotebookLM Research

STEP 0: Bootstrap (MANDATORY — run every time before anything else)

0.1 Check if nlm CLI is installed

0.1 Check if `nlm` CLI is installed