Novel Engine.
The Seven Agents.
Not one general AI — a team of specialists. Each agent has a distinct role, perspective, and editorial voice.
The 14-Phase Pipeline.
A structured build process from concept to published manuscript. Each phase has a designated agent, defined inputs and outputs, and a completion gate that you confirm before the next phase unlocks.
Three user actions per phase: Advance (confirm completion), Done (manually mark complete), Revert (roll back).
Key Features.
Everything you need to build a novel from concept to export-ready manuscript.
Pitch Room
A free brainstorming space where you explore story ideas with Spark before committing to a book. Each pitch conversation gets its own draft folder. Promote to a book, shelve for later, or discard.
Voice Profile System
Before Verity writes a word, you establish a Voice Profile — capturing your sentence rhythm, vocabulary register, dialogue style, emotional temperature, and more through a guided interview. The AI writes in YOUR voice.
Auto-Draft
One click starts a loop where Verity writes one chapter at a time — reading the scene outline, drafting prose, updating the story bible. Continues until every chapter is done. Includes integrated audit/fix passes.
Verity Audit Pipeline
Automated QA after every agent interaction. A Sonnet-powered audit scans for editorial narration, flagged phrases, anti-patterns, and voice drift. Targeted fix pass auto-corrects moderate or heavy issues.
Motif Ledger
Structured tracking for recurring literary elements — motif systems, foreshadowing threads, structural devices, minor character motifs, flagged phrases, and audit logs. Seven-panel tabbed editor.
Revision Queue
Forge produces a revision plan; the queue parses it into structured sessions and executes them. Four modes: manual (approve each gate), auto-approve, auto-skip, or selective. Full task progress tracking.
File Version History
Every file written by an agent or saved by the user is automatically versioned in SQLite. SHA-256 dedup, visual diff viewer, one-click revert, source tracking (user/agent/revert), and version pruning.
Multi-Model Support
Ships with Claude Code CLI as the built-in provider. Add OpenAI-compatible endpoints (OpenAI, Ollama, LM Studio, Groq, etc.) for text completion. Capability-aware model routing with status checking.
Build & Export
Assembles all chapters via Pandoc into Markdown, DOCX, and EPUB. Download All creates a ZIP archive. Catalog Export backs up your entire books library. Copyright page auto-generated from metadata.
Manuscript Import
Import an existing manuscript (.md, .markdown, .txt, or .docx). Chapter detection, preview/rename/merge, then optional multi-agent source document generation (pitch, outline, bible, voice, motif).
Series Bible
Group books into ordered series with a shared story bible. The bible is automatically injected into all 7 creative agents' context — maintaining cross-volume continuity. File-based storage with O(1) reverse-lookup cache.
Series Import
Import multiple manuscripts at once and group them as a series. Auto-detects series name, lets you reorder volumes, skip individual books, and commit to a new or existing series in one batch.
In-App Helper
A floating help assistant (bottom-right corner) powered by a comprehensive 16-section user guide. Answers questions about features, workflows, agents, and troubleshooting. Conversation persists across book switches.
Guided Tours & Tooltips
Three interactive spotlight-based guided tours (Welcome, First Book, Pipeline Intro) with CSS clip-path overlays. Plus contextual tooltips on 14 components — sidebar buttons, pipeline phases, chat controls, and more.
Book Overview Dashboard
Project health at a glance — pipeline progress, word counts per chapter, revision task completion, recent file activity, last agent interaction, and days in progress. All aggregated from the database and filesystem.
Writing Statistics
Token usage over time, per-agent and per-phase breakdowns, word count history, cost estimates, and words-per-chapter charts. Powered by Recharts with interactive visualizations.
Batch Find & Replace
Bulk search-and-replace across all chapter drafts. Literal or regex matching, per-chapter match preview with inline highlighting, selective file application, and automatic version snapshots for safe revert.
Reading Mode
Distraction-free manuscript reading. Assembles all chapter drafts in order into a continuous, formatted view — pure reading without the editing chrome.
Sidebar Bookshelf
A permanently visible bookshelf panel with icon toolbar (New Book, Shelved Pitches, Archives, Series, Import), scrollable book cards with cover images, status badges, and word counts. Series books grouped with collapsible headers.
Five-Tab Files View
Book content organized into five focused tabs: Source documents, Chapters, Agent outputs, Explorer (full directory browser), and Motif Ledger. Plus an inline About.json editor for book metadata.
Getting Started.
Everything you need to start building books.
What You Need
- macOS, Windows, or Linux
- Claude Code CLI — installed, authenticated, and on your PATH
- No API key needed for Claude — the CLI handles its own authentication through your Anthropic subscription
Install Options
Pre-built installers on the Releases page, or build from source with Node.js 20+.
Quick Start
- Install Novel Engine (download or
npm install && npm start) - The onboarding wizard detects your Claude Code CLI
- Set your name, preferred model, and create your Author Profile
- Create a book — or import an existing manuscript
- Start with Spark to pitch your story concept
- Follow the 14-phase pipeline to a finished manuscript
Voice Profile
Verity conducts a guided interview to capture your authentic writing voice — sentence rhythm, vocabulary, dialogue style, emotional temperature. Stored per-book and loaded into every Verity session.
Published Books.
Real novels built with this engine, available on Amazon.