I got tired of ChatGPT 'forgetting' the plot of my stories after 20k words, so I built my own IDE for writing. It's basically if Cursor and Claude Code had a baby, but for novelists.

The Problem: Chat interfaces are linear. They don't have 'project awareness.' You spend 40% of your time pasting context back in.

My Solution (Minotauris): I built an asynchronous agentic workflow.

    Deep Context: It indexes a 'Lore Library' so it knows your characters even 50 chapters later.

    Parallel Agents: I run multiple sub-agents (brainstorming, drafting, refactoring) in parallel on the cloud.

    The Flow: You don't chat. You command. 'Refactor Chapter 4 to match the tone of Chapter 1.' The agent does the work while you sleep.

    Model: Llama 3 8B (via Groq) for speed + selective reasoning calls.

    Architecture: Sub-agent swarm that collaborates with you (it asks for clarification if your command is vague, just like a human editor).