Skip to main content

Integrate with NotebookLM

Goal​

Export the local knowledge graph into a NotebookLM notebook so the captured lessons, ADRs, and patterns can be queried with NotebookLM's AI research surface — including cross-document synthesis and audio overview generation.

Prerequisites​

  • KMGraph initialized with a populated knowledge graph
  • NotebookLM MCP server configured (mcp__notebooklm__* tools available)
  • Authenticated: run nlm login in the terminal if not already authenticated
note

This guide covers exporting the local KG as sources into NotebookLM. NotebookLM becomes a secondary query interface. KMGraph continues to write to local markdown. A NotebookLM-as-primary-store mode is deferred to a future release.

Steps​

1. Create a NotebookLM notebook​

Use mcp__notebooklm__notebook_create to create a notebook named "KMGraph — [Project Name]".

Note the notebook ID.

2. Add lessons as sources​

For each lesson category you want to query, add the files as text sources:

Use mcp__notebooklm__source_add with source_type=file and file_path pointing to
docs/lessons-learned/debugging/my-lesson.md to add it to notebook [ID].

For bulk import, ask the AI assistant to iterate:

Read all .md files in docs/lessons-learned/ and add each as a source to
NotebookLM notebook [ID] using mcp__notebooklm__source_add with source_type=text.
Use the file content as the text field.

3. Add ADRs and patterns​

Read all .md files in docs/decisions/ and add each as a source to notebook [ID].

4. Query the knowledge graph via NotebookLM​

Use mcp__notebooklm__notebook_query with notebook_id=[ID] and query=
"What did we learn about Redis connection issues?"

NotebookLM synthesizes across all sources and cites the specific lessons it used.

5. Generate an audio overview (optional)​

Use mcp__notebooklm__studio_create with artifact_type=audio and notebook_id=[ID].

Then poll for completion:

Use mcp__notebooklm__studio_status to check if the audio overview is ready.

Keeping the notebook current​

After each capture session, add new lessons to the notebook:

Use mcp__notebooklm__source_add to add the new lesson file to notebook [ID].

Or periodically refresh the entire notebook by deleting and re-importing (for large graphs).

Verify​

Use mcp__notebooklm__notebook_query with query="test" to confirm sources are indexed.

The response should cite specific lessons from the knowledge graph.

Next steps​