Predefined AI writing tools, not prompts
Every AI action is a tool that is applied to a part of text: expand, condense, simplify, correct, translate. No prompt engineering, no copy-paste between a chatbot and your document.
AI as functions, not prompts
An online text editor where AI is a button, not a prompt
intell-write is a writing app with AI built into tools. Instead of typing prompts into a chat, you apply tools, such as expand, improve, simplify, and more. Apply them where you are writing and grow a few bullet points into finished prose. You accept or reject every change the AI makes.
See it in action
Watch intell-write at work
A short walkthrough: writing in blocks and applying AI functions, from a few bullet points to finished prose.
Why it works
Three ideas at the core
A writing tool, not a chat window. The AI stays close to your text and always leaves you in control.
Structure first
Write in blocks: bullets, sentences, paragraphs, sections. Blocks assemble the document, so any edits stay local per block, ordered and easy to reason about.
You stay in control
Every suggestion arrives visualizing changes you accept or reject. Insights show who wrote what: you or the AI, word by word.
Designed around the writing, not the model
Most AI writing happens in a separate chat: you describe what you want, paste the result back, and lose the thread of your own document. Intellwrite closes that gap. The AI lives inside a structured editor and acts directly on the blocks you are working on.
By turning common AI operations into named functions, it removes the repetitive prompting that open-ended tools demand. You are not re-explaining your goal each time, you are pressing “improve” on the paragraph in front of you.
The scientific approach
The tool is built on purpose to limit what the AI can do. Instead of an open box where you can type anything, it offers a focused set of writing functions.
This trades functional flexibility for effectiveness: a smaller, well-chosen set of actions makes the tool faster to use and its results easier to predict and control than an unrestricted assistant.
The concept of functional flexibility is taken from our research on text editing with LLMs across conversations, toolbars, and prompts: arXiv:2410.10644.