Skald Scribe

Write complete novels with AI that remembers every chapter

8 followers

Write complete novels with AI that remembers every chapter

8 followers

Skald Scribe turns a story blueprint into a complete novel while maintaining persistent canon, character decisions, timelines, and plot threads across every chapter. Bring your own Claude or Gemini API key and control your generation costs.

Payment Required

Launch tags:Writing•Artificial Intelligence

Launch Team

Framer AI AgentsDesign and publish professional sites with AI

Promoted

Maker

📌

Hey Product Hunt — I'm Brian, the builder behind Skald Scribe. I built Skald Scribe because I kept running into the same problem with AI fiction tools: they could write a good scene. They could write a decent chapter. But they could not hold an entire novel together. By the middle of a book, characters drifted, plot threads disappeared, world rules changed, and the story started forgetting itself. Skald Scribe takes a different approach. Instead of treating a novel like one long chat, it uses a structured manuscript engine built around a Story Blueprint, a Master Ledger, and a six-agent pipeline: Architect plans the chapters. Scribe writes the prose. Inspector checks continuity. Editor polishes the language. Archivist updates the story state. The Blueprint keeps the whole book grounded. The goal is simple: help writers generate full manuscripts while keeping continuity intact across chapters and volumes. Skald Scribe is a bring-your-own-key product. It is $15/month, and authors use their own Gemini or Claude API key, so they pay the AI provider directly instead of paying us a markup. I also wanted people to be able to judge the product before spending anything, so the site includes an uncut demo: a 20-chapter, 45,426-word novel generated in one 72-minute recording, plus the finished novel available to read. This is still early, but it is live, working, and built from the real pain of trying to make AI useful for long-form fiction. I'd love feedback from authors and builders — especially anyone who's tried to push AI past chapter five and watched it fall apart. I'll be here all day answering questions.

Report

2mo ago

How does the persistent canon actually work when the context window gets tight on longer drafts, does it summarize older chapters or hold them in full?

Report

2mo ago

Maker

@orekmeral29034

It does not try to hold the entire manuscript in the model context at once.

Skald Scribe uses the larger context window for the active generation pass, but persistent canon lives outside the prompt in structured story state.

So older chapters are not treated as one giant raw text block forever. The important facts are extracted and carried forward in the Master Ledger: character state, deaths, decisions, locations, world rules, unresolved plot threads, timeline events, and consequences.

When a later chapter is generated, the engine gives the model the current blueprint, the chapter plan, the relevant ledger/canon records, and the recent story context needed for that section. That keeps the working context focused instead of flooding the prompt with every word of the manuscript.

So the answer is: it does not simply summarize the book like a loose recap, and it does not keep every prior chapter in full every time. It keeps structured persistent canon, then pulls the relevant story state forward as needed.

That is the main difference between Skald Scribe and a normal long chat. The manuscript can grow, but the canon memory remains organized outside the chat window.

The 45,000-word demo novel on the site was generated this way — chapter 20 still honoring decisions from chapter 1. Volume 2 generation also carries the same ledger forward, so continuity is designed to hold across books, not just chapters.

Report

2mo ago

Reviews

It does not try to hold the entire manuscript in the model context at once. Skald Scribe uses the larger context window for the active generation pass, but persistent canon lives outside the prompt in structured story state. So older chapters are not treated as one giant raw text block forever. The important facts are extracted and carried forward in the Master Ledger: character state, deaths, decisions, locations, world rules, unresolved plot threads, timeline events, and consequences. When a later chapter is generated, the engine gives the model the current blueprint, the chapter plan, the relevant ledger/canon records, and the recent story context needed for that section. That keeps the working context focused instead of flooding the prompt with every word of the manuscript. So the answer is: it does not simply summarize the book like a loose recap, and it does not keep every prior chapter in full every time. It keeps structured persistent canon, then pulls the relevant story state forward as needed. That is the main difference between Skald Scribe and a normal long chat. The manuscript can grow, but the canon memory remains organized outside the chat window. The 45,000-word demo novel on the site was generated this way — chapter 20 still honoring decisions from chapter 1. Volume 2 generation also carries the same ledger forward, so continuity is designed to hold across books, not just chapters.