Outline
Describe a subject and Mentible drafts a real table of contents — parts, chapters, topics. You set the shape and scope until it is exactly the book you mean.

AI Book Authoring For Self-Learners
Mentible
The content is the commodity — every model already has it. Mentible is where you shape it: outline the book you want, set its scope, and compile a real EPUB or PDF that's exactly what you decided it should be — yours to keep, read offline, and share.
Free to explore · No sign-up · Two complete books inside the app
How it works
Three deliberate steps — that's the whole product.
Describe a subject and Mentible drafts a real table of contents — parts, chapters, topics. You set the shape and scope until it is exactly the book you mean.
Fill it in topic by topic. Each chapter comes back scoped to the level, depth and boundaries you set — with math, diagrams and worked examples. The content is cheap; your spec is the point.
Press publish and Mentible compiles a polished EPUB3 / PDF with a real cover and theme. The finished book is yours — read it offline, keep it, share it.
Why Mentible
The model is the commodity. The shaping — scope, boundaries, structure — is where the value is.
No endless chatbot, no catalogue of canned courses. One focused loop: outline → generate → compile a real artifact.
Works across Anthropic and the OpenAI-compatible providers — OpenAI, DeepSeek, Qwen and more. The model is the commodity; the scoping is the craft.
A managed subscription so you just start writing, or bring your own API key and pay the provider directly. Hybrid by design.
Every book compiles to EPUB3 and PDF with covers, a theme system and a draft → published lifecycle. The book is the product.
KaTeX math, Mermaid diagrams, tables and accessible typesetting render automatically — no formatting busywork.
Self-paced learning for working professionals and the endlessly curious. Go deep, on your own terms — no fluff, no school.
Inside the demo
The demo build is reading-first and fully offline — explore the experience before you author your own.
Transparency
Mentible is AI-powered, and we say so up front. When you generate a book, an LLM writes the draft under the scope and boundaries you set — that's the point, not something we hide. Here's how we keep it honest and verifiable.
Every export carries a Content Trust Manifest — which model produced it, when it was generated, and the settings used. That provenance stays with the file when you export or share it: your proof the material is what it claims to be — disclosed, AI-assisted educational content.
You may have heard that AI models secretly watermark their text. For the models Mentible uses today, mostly not: major text models — including Claude — don't embed a hidden, detectable watermark in the words they generate. The notable exception is Google Gemini, which marks its text output (SynthID), and even that signal can fade when text is edited or recompiled, as it is when Mentible builds your EPUB or PDF. Where a provider does mark its output, we tell you at the point you choose the model.
Plagiarism and AI-detection tools may flag AI-assisted material. With Mentible that isn't something to hide from — your Content Trust Manifest already discloses the model, the date, and how the content was made. Authenticity here means honest provenance, not pretending a human wrote it.
You pick the model; we tell you what that choice means — including any provider marking practices — before you generate. Providers handle this differently, and that difference is shown where you select.
Use the full app right in your browser, peek at the read-only demo, or sideload the Android app and read on your phone. It takes about a minute.
Mentible is in active development — the demo is a reading-first preview; the full authoring app (generation, accounts, managed keys) is on the way.