Write a 700,000-character novel with a team of agents
Claude Code 70 ， 、 ， AI 。 TV Tropes 、 Wiki 、 Agent ， 。 Task: The author used Claude Code to manage a 700,000-character novel as a software project, treating it as a content production system rather than simple text generation. The workflow involved market research on Fanqie Reading, analyzing tropes from TV Tropes, and establishing a file-based structure with world-building, character, and timeline documents. Multiple agents handled drafting, editing, consistency checks, and wiki updates, while the human author guided the process through Socratic dialogue to define core settings and plot arcs.; tools: Claude Code. Human role: The original post does not say how a person stayed involved.; limitations: The approach requires significant upfront setup, including creating detailed style guides, trope libraries, and project directory structures, which may be resource-intensive for smaller projects. The author notes that free-form AI writing fails without strict constraints, and maintaining consistency over 700,000 characters relies heavily on the agent's ability to reference and update external files. Human involvement is critical for high-level plot decisions and verifying that the output aligns with creative intent..
Tools
Published on Agent Work Atlas:
Overview
Claude Code 70 ， 、 ， AI 。 TV Tropes 、 Wiki 、 Agent ， 。
Workflow
The author used Claude Code to manage a 700,000-character novel as a software project, treating it as a content production system rather than simple text generation. The workflow involved market research on Fanqie Reading, analyzing tropes from TV Tropes, and establishing a file-based structure with world-building, character, and timeline documents. Multiple agents handled drafting, editing, consistency checks, and wiki updates, while the human author guided the process through Socratic dialogue to define core settings and plot arcs.
Reported result
The author reports that the method successfully produced a 700,000-character novel, stating that Claude Code could manage long-form consistency if treated as a project engineer. However, the source does not provide measurable metrics on reader reception, sales, or specific quality benchmarks, nor does it detail the exact time or cost invested. The success is attributed to the structured workflow preventing common AI writing failures like plot holes and repetitive pacing.
Human role
The original post does not say how a person stayed involved.
Limitations
The approach requires significant upfront setup, including creating detailed style guides, trope libraries, and project directory structures, which may be resource-intensive for smaller projects. The author notes that free-form AI writing fails without strict constraints, and maintaining consistency over 700,000 characters relies heavily on the agent's ability to reference and update external files. Human involvement is critical for high-level plot decisions and verifying that the output aligns with creative intent.
Try it
Act as a novel project manager using Claude Code. I want to write a [genre] novel of approximately [target word count] characters. First, analyze [platform name] trends to identify [number] popular tropes. Next, create a project directory structure including STORY_BIBLE.md, OUTLINE.md, and a chapters/ folder. Use Socratic questioning to help me define the protagonist's core conflict and the world's magic system. Once the outline is approved, generate a chapter plan for Chapter 1, then write the draft following the style guide in STYLE.md. After drafting, run an editing agent to check for pacing issues and AI-like phrasing. Finally, update the timeline and character files. Please start by asking me three critical questions about the story's premise before generating any content. Confirm the project structure with me before writing the first chapter.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.