notebooklm-py: The 12,000+ Star Python API That Automates NotebookLM

Project Overview

With its powerful document understanding and audio generation capabilities, Google NotebookLM has become a widely popular AI note-taking and research tool. NotebookLM is an AI-powered research tool launched by Google in 2023 (originally called Project Tailwind), built on Google's Gemini large language model. Its core capability allows users to upload various documents (PDFs, web pages, YouTube videos, Google Docs, etc.) and then ask questions, generate summaries, and create notes based on those documents. The Audio Overview feature launched in 2024 catapulted it to fame — it can automatically transform document content into a two-person conversational podcast audio, which quickly went viral thanks to its natural dialogue style.

However, NotebookLM only offers a web interface with no official API, leaving many developers disappointed. Recently, the open-source project notebooklm-py on GitHub has rapidly gained popularity, amassing over 12,000 Stars in a short period and bringing developers full programmatic access to NotebookLM.

Created by developer teng-lin, the project positions itself as an "unofficial Python API and agentic skill." It achieves programmatic access to NotebookLM through reverse engineering — the developer analyzed the HTTP requests and responses between the NotebookLM web application and Google's servers, identified the underlying API endpoints, request parameters, and authentication mechanisms, then simulated these requests with Python code. This approach not only covers all features of the NotebookLM web interface but also exposes some hidden capabilities not available in the Web UI, opening new possibilities for automated workflows and AI Agent integration.

Core Features of notebooklm-py

Complete Python Programming Interface

notebooklm-py provides a full-featured Python API that lets developers perform all NotebookLM operations programmatically:

• Create and manage notebooks
• Upload and process various document sources (PDFs, web pages, text, etc.)
• Generate Audio Overviews
• Conduct conversational interactions with documents
• Extract document summaries and key information

Operations that previously required manual browser interaction can now be batch-automated through Python scripts.

Audio Overview: Automating the Document-to-Podcast Pipeline

NotebookLM's Audio Overview feature is its most iconic capability. The system first uses the Gemini model to transform document content into a structured two-person dialogue script, simulating natural conversation between a host and a guest. It then converts the script into audio using high-quality speech synthesis technology. The generated audio typically runs 5-15 minutes, with natural intonation, filler words, and interactive reactions — sounding almost indistinguishable from a real podcast. With notebooklm-py, developers can batch-trigger this feature to automatically convert large volumes of documents into listenable audio content, particularly suited for scenarios where users need to "listen" to research materials during commutes or workouts.

Three Flexible Usage Methods

The project offers three access methods covering different development scenarios:

1. Python API: Direct integration into existing Python projects with maximum flexibility
2. CLI Command-line Tool: Ideal for quick operations and shell script invocations
3. AI Agent Skill (Agentic Skill): Can serve as a tool plugin for AI agents like Claude Code, Codex, OpenClaw, and others

AI Agent Integration: Letting Agents Control NotebookLM

The project's biggest highlight is its "agentic skill" design. To understand the significance of this design, you first need to understand the concept of AI Agents: AI Agents are AI systems capable of autonomously planning, decision-making, and executing tasks. Unlike traditional single-turn Q&A AI, Agents can decompose complex tasks, invoke external tools, and adjust strategies based on intermediate results. An "Agentic Skill" refers to a specific capability module that an Agent can call — similar to equipping an Agent with different "skill packs."

When notebooklm-py serves as a skill for these Agents, the Agent gains the ability to control NotebookLM. Specifically:

• Claude Code is Anthropic's AI programming assistant that can autonomously write and execute code in the terminal
• OpenAI Codex is OpenAI's programming Agent, skilled at understanding natural language instructions and converting them into code operations
• OpenClaw is another open-source Agent framework

This means you can have these AI programming assistants autonomously operate NotebookLM. For a practical example: have an AI assistant automatically upload a batch of research papers to NotebookLM, generate audio summaries, and extract key findings — the entire workflow requires zero human intervention. The Agent autonomously plans steps, handles exceptions, and decides next actions based on intermediate results.

Community Traction and Project Data

From a technology stack perspective, the project is written in pure Python with a low barrier to installation and use. Current community metrics:

• 12,786 Stars: Reflects strong developer demand for programmatic NotebookLM access
• 1,776 Forks: A large number of developers are actively building upon and customizing this project

This growth rate is quite rare among GitHub open-source projects. For reference, many well-known open-source projects take months or even years to reach these numbers. Similar rapid-growth cases include Auto-GPT in early 2023 (which broke 10,000 Stars in its first week) and various AI Agent frameworks in 2024. This explosive growth typically occurs when a project precisely addresses a common pain point shared by many developers, also reflecting the widespread demand for "API-ification" in the current AI tool ecosystem — developers aren't satisfied with GUI operations and instead want to embed AI capabilities into their automation pipelines. The viral success of notebooklm-py demonstrates that it fills a real and urgent tooling gap.

Typical Use Cases

Batch Document Processing and Knowledge Extraction

Researchers and content creators can write scripts to batch-upload papers and reports to NotebookLM, automatically generating structured notes and audio overviews, dramatically improving literature processing efficiency. For example, a PhD researcher could batch-import hundreds of related papers, have the system automatically generate core finding summaries for each paper, and quickly identify common themes and contradictory viewpoints across papers through conversational queries.

Automated Content Production Pipelines

Integrate NotebookLM into data processing pipelines — for instance, automatically aggregating daily industry news and generating podcast-style audio content, or periodically converting technical documentation into easily digestible audio summaries. Such pipelines can be combined with existing CI/CD systems and task schedulers (like Cron or Airflow) to achieve fully unattended content production.

Building Intelligent Research Assistants

Serving as the "eyes and hands" of AI Agents, giving them document understanding and knowledge management capabilities. Developers can build more complex automated research workflows on this foundation. For example, building an Agent that automatically monitors new papers on arXiv, filters for relevant topics, uploads them to NotebookLM for in-depth analysis, generates research briefings, and sends them to team members via email.

Important Considerations Before Use

As an unofficial project, there are several points to be aware of:

• Interface Stability Risk: The project relies on reverse engineering and may require updates as Google changes its interfaces. Similar historical cases include early unofficial Twitter APIs and various third-party clients for social platforms, all of which experienced frequent adaptation updates when platforms modified their interfaces. Google could modify its internal API structure, add new authentication mechanisms, or implement rate-limiting strategies at any time.
• Compliance Considerations: Users should be mindful of Google's Terms of Service. The legal status of reverse engineering varies across jurisdictions, and legal assessment is recommended before commercial use.
• Long-term Maintenance Depends on the Community: Long-term stability depends on sustained community maintenance. If Google releases an official API, this project may need to reposition itself.

However, given the project's current community activity and contributor count, maintenance and updates should be assured in the short term. The collective wisdom of the open-source community typically enables rapid response and release of patches after interface changes.

Conclusion

The explosive popularity of notebooklm-py confirms a clear trend: developers increasingly need to programmatize and automate AI tools. When a powerful AI product only offers a web interface, the community will spontaneously build programmatic interfaces to unlock its full potential. This phenomenon is particularly prominent in the AI space — from early unofficial ChatGPT APIs to community wrappers for various models, developers have consistently pursued "API-ifying" AI capabilities to integrate them into their own tech stacks.

For developers looking to incorporate Google NotebookLM into automated workflows or add document understanding capabilities to AI Agents, notebooklm-py is currently the most mature open-source solution. As the AI Agent ecosystem continues to evolve, projects like this that transform closed AI products into programmable components will play an increasingly important role in the developer toolchain.