fiction-forge

AI-assisted novel writing toolkit | A Galleys.ai project

Prose pattern scanner, MCP context server, multi-format publisher, and a battle-tested editorial workflow for novel-length manuscripts. Works with Claude Code.

---

AI-generated prose has fingerprints — em-dashes everywhere, show-then-tell constructions, hedging language, premature emotional resolution. At novel length, these patterns compound into something that reads unmistakably machine-generated. fiction-forge detects 24 of these patterns, gives your AI editor real-time access to your story bible, and provides measurable targets for editing. It was built to de-fingerprint a 286,000-word novel — and it worked.

---

What's Included

Tools

• Prose Scanner — Detects 24 overused patterns (em-dashes, similes, AI fingerprints, voice drift) with severity scoring and cluster detection
• MCP Context Server — Gives Claude Code real-time access to your story bible, character profiles, continuity rules, and foreshadowing threads
• Publisher — Compiles markdown chapters into EPUB, PDF, HTML with cover images, part dividers, and professional typography
• Image Generator — Batch DALL-E 3 illustration generator with rate limiting and manifest tracking

Templates

• Story bible, character profiles, foreshadowing ledger, chapter template, editorial notes format, master outline

Presets

• Literary fiction and genre fiction pattern presets with configurable thresholds
• Style profiles (Rothfuss, Sanderson) as examples

Documentation

• Complete workflow guide for parallel-agent editorial passes
• Prose pattern reference, MCP server deep dive, publishing guide
• Lessons learned from a real 286k-word project

Quick Start

git clone https://github.com/geobond13/fiction-forge.git
cd fiction-forge
pip install -r requirements.txt

1. Edit project.yaml with your book's title, author, and part structure
2. Copy templates/chapter.md to book/01_First_Chapter.md and start writing
3. Populate reference/bible.md with your story's voice rules and canon decisions
4. Run python tools/prose_scanner.py --summary to scan for pattern issues
5. Open Claude Code in the project — the MCP server starts automatically

Architecture

project.yaml          Single config file — all tools read from here
     |
     ├── tools/
     │   ├── fiction_mcp.py      MCP server (5 tools for Claude Code)
     │   ├── prose_scanner.py    Pattern detection + severity reports
     │   ├── publish.py          Markdown → EPUB / PDF / HTML
     │   └── generate_images.py  DALL-E 3 batch illustration generator
     │
     ├── presets/
     │   ├── patterns/           Prose pattern definitions (YAML)
     │   └── style_profiles/     Author style references
     │
     ├── reference/              Your story's source of truth
     │   ├── bible.md            Story bible
     │   ├── characters.md       Character profiles
     │   ├── foreshadowing.md    Plant/payoff tracking
     │   └── continuity.md       Timeline and state tracking
     │
     ├── book/                   Your chapters (00_Prologue.md, 01_Title.md, ...)
     ├── templates/              Blank templates for new files
     └── docs/                   Process documentation

The Workflow

The core editorial process uses parallel AI agents coordinated by the prose scanner:

1. Foundation — Build your story bible and outline before writing
2. Draft — Write chapters with MCP context keeping agents grounded
3. Scan — Identify problem chapters by severity tier
4. Fix in waves — Launch parallel agents on non-overlapping files
5. Re-scan — Verify fixes, catch new issues introduced by agents
6. Polish — Sequential voice consistency pass

See docs/workflow.md for the complete process.

Scanner Output

================================================================================
  Prose Pattern Scan — 45 chapters
================================================================================

  Tiers: CRITICAL=2  HIGH=5  MEDIUM=12  LOW=26

  Pattern                   Count   Density    Target
  -------------------------------------------------------
  em_dash                     142     0.87/1k    1.00/1k
  like_a_an                    89     0.55/1k    0.60/1k
  found_myself                 12     0.07/1k    0.15/1k
  show_then_tell                8     0.05/1k    0.10/1k

  Ch    File                           Words  Score Tier     Top issues
  ----  ---------------------------    -----  ----- ----     ----------
  12    12_The_Storm.md                3,421   14.2 CRITICAL em_dash=2.3/1k, 3 clusters
  23    23_Aftermath.md                2,890    8.1 HIGH     show_then_tell=0.4/1k

Case Study: The Third Silence

fiction-forge was developed and battle-tested on The Third Silence, a 286,000-word, 111-chapter fan completion of Patrick Rothfuss's Kingkiller Chronicle.

• 5 complete editorial passes using parallel Claude agents
• 24 chapters expanded, 8 new chapters written with AI assistance
• 60+ chapters modified across the editorial fix plan
• Prose scanner identified and fixed 2,000+ pattern overuses

Documentation

• Getting Started — Setup and first steps
• Workflow Guide — The parallel-agent editorial process
• Prose Patterns — Understanding the scanner
• MCP Server — Context server deep dive
• Story Bible Guide — Building effective reference docs
• Publishing — Multi-format output
• Lessons Learned — Practical insights from a real project

Requirements

• Python 3.11+
• pandoc (for EPUB/HTML generation)
• xelatex (for PDF generation, optional)
• Claude Code (for AI-assisted editing)
• OpenAI API key (for DALL-E image generation, optional)

Contributing

See CONTRIBUTING.md for guidelines on submitting pattern presets, adding tools, and code style.

License

MIT — see LICENSE.

---

A Galleys.ai project | The Third Silence