Claude code can be used to look at a very long book

Install Claude Desktop and get at least a pro account subscription. (low max or high recommended)

<aside> 💡

You do not need to understand GitHub for this next part. A repo is just the main project folder. Create a folder that will contain all the files for the project, and Claude will treat that as the project root.

</aside>

• Open Claude Desktop and go to the Code tab.
• Choose Local.
• Click Select folder.
• Pick the main project folder.
• You will add your manuscript into this hierarchy later

The Initial Instructions that you need to feed Claude Code are in the code blocks below.

What you actually want

You want a local manuscript audit repository that does four things:

1. Ingests one long manuscript into normalized chunks.
2. Runs passes that each produce durable artifacts.
3. Indexes those artifacts so later questions can be answered from the repository instead of from memory.
4. Supports hypothesis work like “can this become 4 books?”, “where does the middle sag?”, “which POV dominates?”, “what scenes are redundant?”

That means your outputs should not just be prose reports. They should also be structured files.

Use both:

• Markdown for human-readable pass reports
• JSON for machine-queryable summaries, scene cards, character appearances, thread tracking, split candidates