No need to design folders and rule files yourself
Chapters, chapter outlines, characters, settings, foreshadowing and storylines each have a place in the editor; there is no AGENTS.md to write and no worry about a rule file exceeding its size cap.
Folio №10 · vs Codex CLI · 2026-10-11
General chat AI, coding agents and overseas writing tools
Can Codex Write Novels? Writing Long Fiction with Codex CLI and How It Differs from Maliang Writer
Codex CLI is OpenAI's open-source terminal agent for local files and commands. Compare its DIY novel setup with Maliang Writer's web editor.
❦ · Verdict · ❦
The short answer
Codex CLI is open source and runs in a local terminal; it reads and writes a whole project and runs scripts, with system-level isolation and approvals. Writing a novel with it, the author builds the folders, writes AGENTS.md, maintains a ledger and uses Git for versions, and views changes as a text diff in the terminal. Maliang builds these into a web editor: a volume-and-chapter table of contents, character cards, story archives and revision comparison you review one by one.
A concrete writing scenario
An author used to working in the terminal pulls the open-source project oh-story-claudecode to a local machine and writes a long mystery with Codex CLI. In the project, settings, outline, body text and tracking are kept separately, the JSON in the tracking directory is the structured source of truth, and the context file has a fixed seven columns and stays under 12KB. He commits to Git before each task and looks at the changes with /diff afterwards. The workflow is rigorous, on the condition that he is comfortable with Git and the terminal and willing to follow the rules the project sets. What Maliang does is turn these rules into structure in the editor, so the author doesn’t touch files.
❦ · Matrix · ❦
10 dimensions compared one by one
On narrow screens, scroll sideways to see all columns.
| Dimension | Maliang Writer | Codex CLI | Notes |
|---|---|---|---|
| Product form | ✅ Web editor plus AI chat, ready to use in the browser | ✅ An open-source (Apache-2.0) local terminal agent for macOS, Linux and Windows; also available on the web, via IDE extensions, and on iOS | |
| Manuscript structure | ✅ Volumes, chapters and chapter outlines are in the table of contents; characters, settings, foreshadowing and storylines are cards | ⚠️ No official material found on novel-specific features; public practice relies on authors building their own folders and AGENTS.md | |
| Writing rules | ✅ Material such as the master outline and style samples is read by AI as needed when writing | ✅ Layered AGENTS.md (global → project root → current directory), with a combined default cap of 32 KiB; /init generates a skeleton | |
| Memory | ✅ After a chapter is written, story archives are organized: character states, relationships, locations, foreshadowing; anything that doesn’t match is listed as pending confirmation; can be turned off | ⚠️ Local Memories are off by default; the official site recommends putting rules that must be followed in AGENTS.md | |
| Change comparison | ✅ Red deletions and green additions on the original in the editor, each of which can be kept or restored | ✅ /diff shows the Git diff (including untracked files) and /review reviews the workspace; in the terminal it is a text diff | |
| Undo and redo | ✅ Undo the current round, branch from before the round, or edit and resend | ⚠️ The official command list has no /undo and recommends making a Git checkpoint before and after tasks; /fork branches a conversation and codex resume reopens a conversation | |
| Long conversations | ✅ The writer agent brings in only the earlier text and characters this chapter needs; when you reach a new volume, old settings are moved into the archive | ✅ /compact summarizes the visible conversation to free up space | |
| Genre reference | ✅ When starting a book, consult the character and structure breakdowns of more than 1,500 popular serials in the public book library; chapter endings are benchmarked against about 56,000 popular chapter endings | ⚠️ No official material found | |
| Price | ✅ Free sign-up to try; credits are deducted by usage | ✅ Included in ChatGPT plans (Free, Go $8, Plus $20, Pro from $100), or pay by usage with an API key (checked 2026-10) | |
| Regions | ✅ A Chinese-language web product for Chinese-language authors | ⚠️ The OpenAI Help Center says accessing from an unsupported country or region may lead to account suspension; ChatGPT’s official supported-country list does not include Mainland China |
✅ Confirmed in official material or in the product ⚠️ Partly supported, or not stated in official material ❌ Officially not supported · Competitor information comes from their public material.
How Maliang Writer handles these problems
Story archives organized after each chapter
Setting cards are edited as you ask; after a chapter is written and this stretch of writing ends, character states, relationships, locations and foreshadowing are organized in the background, and lists anything that doesn’t match as pending confirmation. It can be turned off.
Revision comparison in the editor
When AI edits the body text, red deletions and green additions are marked on the original, and each can be kept or restored; if you are unhappy with a whole round, you can undo the current round, branch from before it, or edit and resend, with no Git checkpoints to create yourself.
- Works in the browser: No npm or Homebrew to install and no terminal to configure; open the browser and write, and new chapters appear directly in the left-hand table of contents.
- Genre reference for Chinese web novels: When starting a book, it consults the character and structure breakdowns of more than 1,500 popular serials; the writer agent writes the body text from this book’s earlier text and style samples, and each chapter comes with a chapter-ending check, benchmarked against about 56,000 popular chapter endings.
What the author has to do when writing a long novel directly with Codex CLI
- The folders and AGENTS.md are yours to write: In public practice, Claude-Code-Novel-Writer uses AGENTS.md as the general standard, adapted for both Claude Code and Codex; the author has to understand these rules to change them. AGENTS.md has a combined default cap of 32 KiB, so when there are many rules the author has to trim them.
- Versions are yours to manage with Git: The official site recommends Git checkpoints before and after tasks; GitHub issue #11626 notes that rewinding with Esc rewinds only the conversation history and not file changes (still open as of 2026-10).
- The setting ledger is yours to maintain: To stop characters from knowing the answer too early, oh-story-claudecode records “the author’s truth” and “what the reader knows” separately; a ledger like this has to be kept up to date by the author according to the project’s rules.
- Changes are viewed in the terminal: /diff shows a Git text diff. For authors used to reading typeset body text, reading a terminal diff takes getting used to.
What Codex CLI does well
Codex CLI is open source and runs locally, can look at files, edit files and run commands; in the official words, “Inspect code, make changes, run commands, and automate repeatable work without leaving your terminal”. It has system-level isolation and approvals, network access off by default and telemetry off by default; layered AGENTS.md lets rules be subdivided by directory. Nym Coy, a writer who uses VS Code, observed that “Codex defaults to writing a script”, which is handy for batch processing a manuscript.
Every product suits some scenarios; the advice here covers only one: serialized long-form fiction in Chinese.
Maliang Writer's real interface
❦ · Evidence · ❦
Sources and check dates
- Codex CLI repository: OpenAI’s local coding agent, open source under Apache-2.0, for macOS, Linux and Windows. View →
- About AGENTS.md: Read level by level from global down to the current directory, with a combined default cap of 32 KiB; /init generates a skeleton. View →
- Codex CLI version advice: The CLI docs recommend making Git checkpoints before and after tasks so you can roll back. View →
- GitHub issue #11626: “Esc rewind only rewinds conversation/history … but does not revert code changes”, raised on 2026-02-12 and still open. View →
- Codex pricing: Included in ChatGPT plans, or billed at API prices with an API key. View →
- oh-story-claudecode（GitHub）: Can be installed into Claude Code, Codex CLI and OpenCode; the tracking JSON is the structured source of truth, and context files are capped at 12KB. View →
- Using VS Code to Write Fiction with AI: One file per scene, with a script to merge the chapters; mentions “Codex defaults to writing a script”, and that for non-programmers VS Code “sounds intimidating”. View →
- Maliang: revision comparison and undo: AI changes are reviewed one by one in the editor; you can undo the current round, branch, or edit and resend. View →
Try four real writing tasks before you decide
For this page, record in particular how much work the author has to do to prepare folders, writing rules, material updates and revision checks. Don't compare only the prose of a single generation, and don't assume in advance that folder-based tools must be harder to use.
Below is a same-task trial checklist, not a finished comparison experiment. You can use the original teaching material “Jiehuo City” or a manuscript you have the right to handle; record the preparation, the output and the author’s edits for each tool.
Start a new book
Give both tools the same protagonist, core rules and reading preferences; compare the opening directions first, then organize the master outline, the first volume outline and the outlines for the first three chapters.
Check: Can it tell the roles of the three outline levels apart? Can the author confirm a plan before it continues? Were settings that hadn't been settled yet quietly written in as fact?
Use the same book-opening kitContinue an existing twenty chapters
Use the same old manuscript you have the right to handle, specify how the last chapter ends and the goal of the next chapter; try writing one scene first, without changing the previous twenty chapters.
Check: What material did you prepare? Did character positions, items and open threads carry over? How are candidate drafts told apart from adopted ones?
See the import-and-continue task checklistChange a setting that has already appeared
Pick a real rule that affects earlier chapters; first ask it to list the affected chapters and the evidence, then let the author choose the scope of the change.
Check: Does it say which parts it did not check? Can you review the changes first? Did it wrongly edit text outside the task or write an unadopted plan into the material?
See how to change a settingTake your work and material with you
Actually try keeping the text, the order of volumes and chapters, character profiles and world rules, and check whether the output is suitable for further editing; list what cannot be taken along.
Check: Are the format and order recognizable? What material is inside the platform account, what is in local files, and what is in the export? Do not infer usability from a marketing page alone.
Copy the task and result record cardRecord the models, plans or account conditions you actually used, and the time spent preparing materials and editing the text. Leave blank anything you have not tried or cannot confirm; do not let a feature name or a single generation stand in for the full experience.
Choose by your situation
- If You are used to the terminal and Git and want to process your manuscript in bulk with scripts → choose Codex CLI, because Codex CLI can run commands and write scripts, which suits automated processing..
- If You need open source, local execution and permission control → choose Codex CLI, because Codex CLI is open source, by default only allows changes to workspace files and keeps network access off..
- If You don’t want to write in a terminal and want to write Chinese web novels directly → choose Maliang Writer, because The table of contents, cards, archives and revision comparison are all in the web editor..
- If You already have a ChatGPT plan and want to try both → use both, because Codex is already included in ChatGPT plans; try the first few chapters of the same book on each and compare the effort of maintaining a ledger and reading changes..
FAQ
Can Codex write novels?
Yes. Codex CLI reads and writes files and runs commands locally, and open-source projects give it novel folders and AGENTS.md rules. It is a programming tool with no novel-specific features, so the author has to build the folders, rules and ledger.
What do I need to prepare to write a novel with Codex CLI?
Install Codex CLI (via npm, Homebrew and so on), sign in with a ChatGPT account or configure an API key, and prepare a manuscript folder and an AGENTS.md; Git is recommended for versions. You can also use an open-source novel project on GitHub as a starting point.
What should AGENTS.md for writing a novel contain?
Commonly: style, point of view, forms of address for characters, which file is the source of truth, and which ledgers to update after each chapter. Note that the AGENTS.md files Codex reads have a combined default cap of 32 KiB.
Can Codex CLI undo changes?
The official command list has no /undo, and the CLI docs recommend making Git checkpoints before and after tasks; an author pointed out in GitHub issue #11626 that rewinding with Esc rewinds only the conversation and not file changes. Maliang can undo the current round, restoring the manuscript to just before that message was sent (manual edits made afterwards will also be reverted, and credits already deducted are not refunded).
Does writing a novel with Codex cost money?
Codex is included in ChatGPT plans; the Free tier has a limited quota and Plus is $20/month; you can also pay by usage with an API key (checked 2026-10).
Is Codex the same as Novelcrafter’s Codex?
No. Codex CLI is OpenAI’s terminal coding agent; Novelcrafter’s Codex is that writing software’s story bible.
Can I use Codex to write novels in China?
The OpenAI Help Center says accessing from an unsupported country or region may lead to account suspension; ChatGPT’s official supported-country list does not include Mainland China.
Will Codex forget settings when writing a novel?
Codex’s local memory is off by default, and the official site recommends putting rules that must be followed in AGENTS.md. Character states and foreshadowing in a long novel usually rely on ledger files the author designs. In Maliang, story archives are organized in the background after a chapter is written and lists anything that doesn’t match as pending confirmation.
❦ · Continued · ❦
Related comparisons
Can Claude Code Write Novels? How Public Practice Works and How It Differs from Maliang Writer
vs Claude Code · updated 2026-10-11
Claude Code can write novels, and it is very capable: it reads and writes a whole book’s folder, checks continuity across files and makes batch edits. In public practice, authors design the folders, write CLAUDE.md, maintain a setting ledger and work in a terminal or editor plugin, which suits authors who are comfortable on the command line. Maliang builds the table of contents, character cards, story archives and revision comparison into a web editor, and the author starts writing from the chat.
Read the comparison →How Good Is Cursor for Writing Novels? Public Practice and How It Differs from Maliang Writer
vs Cursor · updated 2026-10-11
Cursor can write novels: the editor lets you see files and live diffs, the Rules system is detailed, and the choice of models is wide. In public practice, authors build their own chapter folders, write rules and maintain settings and a ledger, and once there are many chapters they also have to deal with file ordering and merging. Maliang builds the volume-and-chapter table of contents, character cards, story archives and revision comparison into a web editor.
Read the comparison →Writing Novels with ChatGPT vs Maliang Writer: How Projects and a Novel Editor Divide the Work in a Long Serial
vs ChatGPT · updated 2026-10-11
ChatGPT suits brainstorming, drafting passages and polishing sentences. Projects keeps story files and instructions together, and Memory remembers preferences. In a long serial, how chapters are ordered, where to edit when a setting changes and what this draft changed are still for the author to arrange. Maliang puts volumes and chapters, character cards, story archives and revision comparison in one editor.
Read the comparison →Coding Agents vs Novel Writing Software: The Trade-offs of Writing Long Fiction with Code Tools
vs Coding agents · updated 2026-10-11
Coding agents are very capable: they read and write across files, make batch edits and run scripts, and technical authors can build very flexible writing workflows. The price is that the folders, rule files, setting ledger and version management all have to be designed and maintained yourself, and the interface is aimed at programmers. Maliang turns these into the cards, table of contents and editor that authors are used to.
Read the comparison →How to Manage Novel Settings: Folders + Markdown, or Character Cards and Story Archives?
vs Folder + Markdown approach · updated 2026-10-11
Folders plus Markdown are flexible and transparent, with the files in your own hands, which suits authors who like to define their own formats; the drawback is that once there are many files, which file is the latest and which other places need syncing after one edit are all yours to remember. Character cards plus story archives fix each character and setting as a card and organize states automatically after a chapter is written, which suits authors who don’t want to maintain files. Maliang has both: characters and settings are cards, the master outline is stored separately, and style samples and book-opening material are documents.
Read the comparison →Try one chapter of the book you're writing
After signing up, create a book or import existing chapters, tell the AI which chapter to continue, then look at what it wrote, what it changed and what it recorded in the archive, and decide whether to move your long novel over.