Why I Built Codiko
This project started because I wanted to write a book.
Not a simple book — a sprawling, multi-timeline, world-heavy piece of fiction with dozens of characters, branching storylines, and rules I needed to keep straight across hundreds of pages. The kind of story where you're constantly flipping back to check "wait, did I already establish that?" and "which chapter did that conversation happen in?"
I tried the tools that exist. They didn't work.
The Problem with Existing Tools
Scrivener is the standard, and for good reason — the binder system is solid for organising manuscripts. But it has no timeline, no character relationship tracking, no AI, and the interface feels like it hasn't changed in a decade. For a straightforward novel, it's fine. For a complex world with branching timelines, it's not enough.
Google Docs and Word are great for writing prose, terrible for managing anything around it. I ended up with a manuscript in one place, character notes in another, a timeline spreadsheet somewhere else, and a folder of reference documents that I never remembered to update.
Notion can theoretically do everything, but building a writing system from scratch in a general-purpose tool is exhausting. And none of it talks to the AI — you still end up copy-pasting chapters into ChatGPT and losing context every time.
ChatGPT/Claude are powerful, but they forgets a lot. Every conversation starts from near zero. You can paste in context, but there's a limit, and managing that context is another job on top of writing.
What I Wanted
I wanted a single workspace where:
- My manuscript, characters, locations, and timelines all live together
- The AI can read and search across all of it
- I can ask a question about chapter 3 while writing chapter 30 and get an accurate answer
- The AI suggests edits I can review before accepting — not auto-generated prose I didn't ask for
- My data stays on my machine
- I'm not paying a subscription forever
Nothing like that existed. So I started building it.
Building Codiko
The app is built with Tauri (Rust backend, React frontend), which means it's a proper native desktop app — fast, lightweight, and your data stays local. The AI connects to your own API key (OpenAI, Anthropic, or local models via Ollama), so there's no data harvesting and no markup on AI costs.
The hardest part was the search and indexing system — the thing that lets the AI search your manuscript and canon before responding. Getting it to return genuinely relevant context, not just keyword matches, took a lot of iteration. But it's what makes the whole thing work.
Where It's Going
Codiko is launching on Mac and Windows. The roadmap includes a lot more — better AI tools, more export options, community templates — but the core is solid and it's ready for writers to use.
If you're working on a complex story and you've felt the friction of tools that don't keep up with what's in your head, give it a try. I'd love to hear what you think.