NeuroBook

English

简体中文

English

简体中文

Appearance

NeuroBookLong-form writing deserves an IDE

Almost everyone has a novel in them, and most of them die halfway. Not from a lack of talent — from a lack of engineering. Worldbuilding, setups, prose and AI collaboration, all in one workspace you can actually see.

Download and Install
Write Your First Book
What Is NeuroBook
简体中文

🌍 Your world stops contradicting itself

The arm that got severed last volume does not grow back on its own. World Engine records world state as a timeline of slices, so the state of any character, faction or inventory count at any moment is computed rather than remembered — which means it never drifts. Adding backstory is just inserting a slice at the right moment, so flashbacks and revelations come for free.

🧵 Setups you plant actually get paid off

Planted a gun in chapter 3 and still not fired by chapter 200? The promise ledger tracks every setup as a debt you owe the reader: planted, advanced, fulfilled — all auditable, and it flows into the writing brief when you reach the target chapter. Want a romance beat every few chapters? Track that too, and it will tell you it has been thirty chapters.

🗂️ The manuscript is your own files

Worldbuilding lives in lorebook/, prose lives in manuscript/ — plain local Markdown files plus one project SQLite database. Any editor can open them, and moving house is copying a directory. No cloud lock-in and no export feature needed, because they were files all along.

✍️ A whole AI writing crew

Not one chat box doing everything. leader plans and dispatches, writer writes prose, retrieval looks up your worldbuilding, researcher checks external sources — so numbers are not invented (the engine has them on the books) and facts are not guessed (it goes and checks). Discuss mode gives ideas without touching the manuscript; plan mode proposes first and executes only after you approve.

🧹 Strip out the AI smell

Lint your prose the way eslint lints code. llmlint ships 340 rules covering filler words, mechanical transitions, formulaic parallelism, hollow summaries and other AI tells. Static rules scan a full manuscript in seconds, and mechanical issues can be fixed automatically. It works as a polishing skill inside the editor and as a standalone CLI.

🧭 An assistant that has read the manual

You do not have to master a complex tool. The built-in assistant has read the entire documentation set — ask it "what should I do first on a new book" or "how do I register a setup" and it will teach you, or just do it for you. The only skill required is typing.

💻 Runs on your own machine

Unzip and run on Windows; container or Bun on Linux and macOS. The database is a local SQLite file, you bring your own model provider and API key, and token spend is broken out by input / output / cache so you can see exactly what a chapter cost.

Where to Start ​

Not installed yet? Read Quick Start — Windows users unzip and run, and you can be writing in five minutes. To choose a deployment method or install on a server, read Deployment.

Already installed? Go to From Your First Book to Chapter Three: create a project, build the worldbook, initialize the world engine, and write the first three chapters.

Want to understand it first? Read the Introduction, or go straight to the four core capabilities: World Engine, Plot Workbench, Markdown Studio, llmlint.

Documentation Map ​

  • Introduction: what NeuroBook is, who it is for, and how it differs from an ordinary AI chat tool.
  • Quick Start: download, launch, configure a model — the shortest path.
  • Tutorials: from your first project to chapter three, in six sessions.
  • Core Capabilities: World Engine, Plot Workbench, Markdown Studio and llmlint in depth.
  • Guides: settings, themes, change history, account and cloud backup.
  • Agent: the mental model behind agents, sessions, profiles, skills, workflows and the three modes.
  • Profile: what each built-in profile does and where its boundaries are.
  • Profile TSX: write your own agent profile.
  • Deployment and Operations: installation, starting and stopping, where your data lives, privacy boundaries.
  • Design Notes: why writing is split across multiple agents.

About AI Roleplay ​

Early versions of NeuroBook shipped AI roleplay (RP) and world simulation modules. The current version has removed the RP entry points from the regular interface while it is redesigned to the standard set by writing mode; there is no timeline yet. The related profiles, data structures and historical material all remain in the codebase.

SillyTavern character card import still works: an inspect → unpack → import three-step flow that archives the original card and its worldbook intact and moves stable settings into the worldbook. What you get afterwards is material for novel writing — it does not start an RP session. See Import a Character Card.

More Entry Points ​