anything-to-notebooklm

This MCP server provides automated content extraction and conversion workflows that feed content into Google NotebookLM via Claude Code Skill. It automatically detects the input type (article URLs, videos, PDFs, EPUBs, etc.), uploads content to NotebookLM, and generates output formats such as podcasts, slides, mind maps, quizzes, or reports. You can trigger multi-source processing from natural language prompts like “把这篇文章生成播客” or “把这些内容做成报告” and the MCP pipeline will handle extraction, transformation, and delivery locally when possible. The server is designed to work with the Claude Code Skill ecosystem and relies on the wexin-read-mcp component to fetch content (e.g., 微信公众号) and NotebookLM for AI-driven generation. Typical outputs include audio (.mp3), PDFs, Markdown, or JSON artefacts representing the generated structure.

Prerequisites:

• Python 3.9+ installed
• Git installed
• Access to Claude Skills directory (as used by your environment)

Step-by-step:

1. Clone the project into your Claude skills directory:

# 1. Clone to Claude skills directory
cd ~/.claude/skills/
git clone https://github.com/joeseesun/anything-to-notebooklm
cd anything-to-notebooklm

2. Install dependencies using the provided script:

# 2. Install dependencies
./install.sh

3. Configure MCP within Claude Code according to prompts in the installation flow (no manual config required in this README beyond the steps above).

4. Restart Claude Code to apply the MCP changes.

First-time usage checks:

# NotebookLM authentication (only once)
notebooklm login
notebooklm list

# Optional environment checks
./check_env.py

To verify the MCP server specifically, you can run the MCP server script directly (from the path shown in the README TROUBLESHOOTING section):

python ~/.claude/skills/anything-to-notebooklm/wexin-read-mcp/src/server.py

Tips and caveats:

• The MCP workflow relies on NotebookLM; ensure you have completed the NotebookLM login (notebooklm login) before triggering generation.
• The project supports multiple content sources (微信公众号, YouTube subtitles, PDFs, EPUBs, web pages, and more). If a source is not natively supported, the MCP may route to the general content handling flow.
• Environment variables shown in the mcp_config field are placeholders. Adjust them according to your deployment environment and NotebookLM access method.
• If you encounter “MCP tool not found” errors, use the troubleshooting commands in the README to reinstall dependencies or re-run installation scripts.
• This MCP is Python-based; there is no npm package for the server itself.
• Security: be mindful of sensitive data in source content; the tool emphasizes local-first processing where possible.