Claude Code  ·  Plugin

notebooklm-memory
gives Claude a memory that outlives the chat.

A Claude Code plugin that wires Claude to Google NotebookLM as permanent storage. It remembers your decisions across sessions, and it checks a curated library of engineering standards before it writes code — so the code comes back consistent, not improvised.

v1.0.0 MIT Python 3.10+ free · Google account github.com/hanialnaber/notebooklm-memory
install · in Claude Code
# 1 · add the marketplace, then install the plugin
/plugin marketplace add hanialnaber/notebooklm-memory
/plugin install notebooklm-memory@notebooklm-memory-marketplace

# 2 · one-time setup (installs the CLI, logs you in, makes your AI Brain)
/notebooklm-memory:memory-setup

Two lines to install, one command to set up. Everything below explains what you just got.

The idea

Two notebooks, two jobs

The plugin manages two NotebookLM notebooks and teaches Claude when to read and write each one. You never touch NotebookLM directly — Claude does.

🧠

AI Brain

Your personal memory. Say wrap up and Claude writes the session's decisions, work, and open threads here. Next time, it reads them back before answering “what did we decide about X?”

📐

SDLC Standards Brain

A curated library of ~50 engineering standards — style guides, OWASP, testing, architecture, git, SRE, docs. Claude consults it before writing or reviewing code, and cites what it used.

What's inside

One skill, four commands, a builder

ComponentWhat it does
skill · notebooklm-memoryRuns automatically on intent — consults the standards before coding, and saves / recalls sessions. No command needed.
/…:memory-setupOne-time install: checks Python, installs the CLI, logs you in (bundled Chromium), installs the skill, creates your AI Brain.
/…:sdlc-buildBuilds or refreshes the SDLC Standards Brain from the bundled 50-source list. Idempotent — safe to re-run.
/…:sdlc-consultAsk the standards notebook a direct question and get a cited answer.
/…:wrap-upSummarize the current session and save it into the AI Brain.
build_sdlc_notebook.pyCross-platform builder: create/reuse notebook → add sources → poll until ready → report. Reads sdlc-sources.txt.

Commands are namespaced /notebooklm-memory:<name>. Notebooks are resolved by title at runtime, so the plugin works for anyone who installs it — nothing is hard-coded to one machine.

Install & set up

From zero to a working memory

  1. Check the one prerequisite

    You need Python 3.10+ and a free Google account. Verify Python:

    python --version   # need 3.10 or newer
  2. Add the marketplace & install

    In Claude Code:

    /plugin marketplace add hanialnaber/notebooklm-memory
    /plugin install notebooklm-memory@notebooklm-memory-marketplace
  3. Run one-time setup

    This installs the notebooklm-py CLI, opens a browser to sign in, installs the skill, and creates your AI Brain. When it pauses, finish the Google sign-in.

    /notebooklm-memory:memory-setup
  4. Build the standards library

    Adds ~50 curated standards and waits until each is ingested. Takes a few minutes.

    /notebooklm-memory:sdlc-build

⚠ Login gotcha

Setup logs in with the bundled Chromium on purpose. Never sign in with --browser chrome or --browser msedge while that browser is already open — the OS hands off to the running window and login fails with “Opening in existing browser session.”

Using it day to day

Mostly, you just work

The skill fires on its own. These commands are there when you want to be explicit.

…before Claude writes codeautomatic

The skill queries the standards notebook and applies the relevant rule — “per OWASP ASVS 5.0”, “per the Google TypeScript style guide” — instead of guessing. You'll see it cite the source.

/notebooklm-memory:sdlc-consult "REST error response conventions"on demand

Ask the standards library a direct question and get a cited answer, any time.

wrap upjust say it

Claude summarizes the session — decisions, what was done, open threads — and saves it into your AI Brain. Nothing is lost between chats.

"what did we decide about the auth flow?"recall

Claude reads the AI Brain before answering, so past decisions carry forward across sessions.

Make it yours

Customize the standards

The 50 sources live in scripts/sdlc-sources.txt — one URL per line, # for comments. Edit it for your stack, cloud, or branching model and re-run /notebooklm-memory:sdlc-build; the builder reuses the notebook and skips sources already present.

Two high-value sources the builder can't fetch for you — it reminds you each run:

Add these yourself

Your team's conventions doc — the single most valuable source, because it's the one thing that's actually about your code. And books you own (copyrighted) — add both as file sources from your own copies.