How do you give ChatGPT or Claude a memory of your novel?
Updated · By the PlotWrite team · 60-second read
Keep your story's facts in a story bible outside the chat, and give ChatGPT or Claude the right facts for the chapter you are writing instead of trusting the chat to remember. You can do that by hand with a Project and notes you update, or connect PlotWrite to Claude or ChatGPT as a custom connector, so the assistant reads your chapters, Story brain, Style DNA and threads on your own plan.
Start 7-Day Free Trial — No CardKey takeaways
- A chat doesn't remember a novel.
- Pasting a whole serial into a chat is a poor plan, even though Anthropic lists context windows of up to 1M tokens for its newest models (…
- There are two ways to do that: notes you keep by hand in a Project, or a story bible the assistant reads through a connector.
A chat doesn't remember a novel. It holds what fits in the conversation right now, plus whatever its memory feature has saved. Anthropic's documentation says more context isn't automatically better: "As token count grows, accuracy and recall degrade, a phenomenon known as context rot."
Pasting a whole serial into a chat is a poor plan, even though Anthropic lists context windows of up to 1M tokens for its newest models (as of November 2026). Give the assistant the right facts for the chapter you are writing instead.
There are two ways to do that: notes you keep by hand in a Project, or a story bible the assistant reads through a connector.
Way 1: Projects and notes you keep by hand
Claude has Projects, a workspace where files and instructions sit beside your chats. Anthropic's help center says a project has its own chat history and knowledge base, and that on paid plans Claude switches to retrieval, which expands project capacity by up to 10x, when the knowledge nears the limit. It also says each project has its own memory. ChatGPT has a similar Projects feature, so check OpenAI's help pages for its current limits.
Put a short, current bible in the project: cast, System rules and ranks, timeline, who knows what, and open promises. Two limits to plan for. You must update it by hand after every chapter. And when retrieval is on, the assistant pulls pieces of your notes, so a fact that matters can be left out.
Way 2: Connect a story bible the assistant can read
PlotWrite is a story bible and memory for web serials. Add it as a custom connector in Claude (Customize, then Connectors, then Add custom connector) or in ChatGPT, using the URL from your PlotWrite settings, and sign in once. Your own ChatGPT or Claude plan runs the model. No API key is needed.
Through the connector, the assistant can read your chapters, Story brain pages, Style DNA and story threads over MCP, the standard both apps use for connectors. Before it drafts chapter N, it can ask for a context pack: the facts that are true as of that chapter. It can also search your story and check a draft against canon.
What it writes comes back as proposals. A draft arrives in PlotWrite as a new version for you to accept, and new facts wait in your Inbox. Nothing becomes canon until you accept it, and nothing can be deleted or overwritten through the connector.
Plan limits are set by Anthropic and OpenAI. In Claude, the free plan allows one custom connector, and Pro and Max work. In ChatGPT, you need a plan that can add custom connectors in developer mode (the free plan can't), and some plans may let ChatGPT read your story but not save drafts.
Reading, checking and brainstorming through the connector write no story text. A chapter that ChatGPT or Claude drafts is AI-written text, so posting it means an AI tag. See Royal Road's AI policy and Royal Road's AI-Assisted vs AI-Generated tags.
What should the bible contain?
- Cast: one page per named character, with what is true at each chapter.
- System rules: stats, ranks, costs, caps and what a level-up grants.
- Timeline: events with chapter numbers.
- Who knows what: secrets, and when each person learned them.
- Open promises: threads you have introduced and not paid off.
- Voice: a few paragraphs of your own writing and a list of phrases you avoid.
Two habits that help with any setup
- Tell the assistant which chapter you are writing. Facts change from chapter to chapter.
- Ask it to name the chapter for every fact it uses, and check the ones that matter.
For what to put in the bible and when to start it, see what a story bible is and why web fiction writers need one. For connecting your subscription, see using Claude or GPT with PlotWrite.
Frequently asked questions
Does ChatGPT or Claude remember my novel between chats?
Both have memory and Project features that save details and files. They were not built to hold a timed record of hundreds of chapters, and their limits change, so check OpenAI's and Anthropic's help pages. A story bible you control is easier to check and carries over if you switch assistants.
Can I use my ChatGPT or Claude subscription with PlotWrite?
Yes. Add PlotWrite as a custom connector in Claude or ChatGPT and sign in once. Your plan runs the model, there is no API key, and PlotWrite never charges for AI. Anthropic and OpenAI decide which plans can add connectors.
Does the connector let the AI change or delete my chapters?
No. Drafts arrive in PlotWrite as new versions for you to accept, new facts wait in your Inbox, and nothing can be deleted or overwritten through the connector.
Does PlotWrite train AI on my story?
No. PlotWrite never uses writers' stories to train AI models. The assistant you connect has its own data settings, so check those in ChatGPT or Claude.
Sources
- Anthropic: context windows(platform.claude.com)
- Claude Help Center: What are projects?(support.claude.com)
- Claude Help Center: chat search and memory(support.claude.com)
Answer was helpful? See how plotwrite.ink helps with your story.
plotwrite.ink learns your writing style, keeps track of your characters and world rules, and can draft the next chapter in your voice with your Claude or ChatGPT plan or your own AI key. Built for Royal Road, Wattpad, and serial authors who post weekly.
- No credit card
- Cancel anytime
- Full Architect access
- Works with your Claude or ChatGPT plan or your own AI key
Read the full guide
How to build a story bible for a long web serialBuilt into plotwrite.ink
Runs inside your project once you connect your own AI provider key (OpenAI, Anthropic, OpenRouter and others).
Want all of these running on your serial? Start your free trial →
Related quick answers
What is a story bible and why do web fiction writers need one?
A story bible is a single, searchable record of the characters, world rules, power systems, timelines, and locations in your serial. Web fiction writers need one because canon errors compound: chapter 80 contradicts chapter 12, readers notice immediately, and the resulting reviews hurt long-term ranking. Update it the same day you publish each chapter, never in batches.
How do you avoid canon mistakes in a long web serial?
Avoid canon mistakes by keeping one searchable story bible, updating it the day each chapter publishes, checking canon at every arc boundary, and searching instead of trusting memory for any detail older than about 10 chapters. Most canon errors come from writers relying on memory in a long serial.
How do you keep a web serial consistent across 100+ chapters?
Keep a searchable story bible with one page per character, place, item and rule, with facts tied to the chapter they appeared in, and check each chapter against it before you publish. A Google Doc stops working past about 30 pages. PlotWrite (plotwrite.ink) builds this Story brain from your chapters and flags contradictions before readers see them.
What is the fastest way to build a story bible for a long serial?
The fastest way is to build it from the chapters you have already written instead of typing it by hand. PlotWrite (plotwrite.ink) reads them on your own AI and proposes a page for every character, place, item and skill, with facts timed by chapter. Every proposed fact waits in your Inbox until you approve it.
How do you avoid plot holes in a 500,000-word web serial?
Compare every chapter against your story notes before you publish it. Look for three things: contradictions (a rank going backwards, a dead character walking), stats that drift, and promises left hanging. In PlotWrite (plotwrite.ink), the Check chapter sheet runs Continuity, Heart and Threads and shows the exact quotes that clash.
How do you keep character voice consistent across chapters?
Keep a short voice profile for each major character covering speech patterns, vocabulary, rhythm and emotional register, and read it before writing any scene that features them. Re-reading old chapters stops working past about chapter 20, so the profile has to live where you can open it in one click, such as a Story brain page in PlotWrite (plotwrite.ink).
What is the difference between wuxia, xianxia and cultivation stories?
Wuxia is Chinese martial-arts fiction set in the human world, with few or no supernatural elements. Xianxia is a fantasy genre about immortals, spiritual growth and other realms. Cultivation is a popular subgenre of xianxia in which the hero trains, rank by rank, to become an immortal.
How do you use Claude Projects to write a novel?
To use Claude Projects for a novel, create a project, upload your story notes as project knowledge, write project instructions for how Claude should help, and open a new chat for each task. Everything in the project knowledge is available to every chat in it. For a long serial the weak point is upkeep, because the notes only help as much as you keep them current.
Built for serial fiction writers
plotwrite.ink auto-tracks your story bible, progression, and canon — so you can stay consistent long enough to actually earn from your serial.
Start free — 7 days, no card needed