Markelangelo is a desktop markdown editor with a serious AI layer built in. The AI works inside your document, your prompts become reusable tools, and nothing leaves your machine unless you say so.
The founders of Markelangelo are veteran agentic architects and developers. When the first large language models arrived, their workflow changed overnight: AI helpers slipped into everything they did, and it was wonderful. The trouble was everything around it. The endless cutting and pasting. Never quite remembering where the good, hard-won prompt had got to. Being buried under a pile of well-meaning tools and services. And that daily hunt through a wall of browser tabs for the one you actually wanted.
So the question answered itself: what did we actually want? We knew, so we built it. Markelangelo began as a frankly selfish project, a way to optimise our own days while keeping all the pretty things we had grown to rely on. We have used it every single day since, whittling and fine-tuning it into what you see now.
We are far from finished. The plan is to grow Markelangelo into a safe pair of hands for all document generation and manipulation. And if you wondered why it is markdown-native from the ground up, that was deliberate: markdown gives you the most flexibility in a world where nearly every AI tool now generates and consumes it natively. When you do need to leave that world, Rich Switch turns your draft into a real Word document and opens it in Markelangelo's own word processor, ribbon and all, with every AI feature still to hand. So why not take it for a spin, and see how it pulls your scattered activities back into one place.
In case you were wondering: we never share your details with anyone, and we will only ever contact you when we have something material to announce. Have a great day.
You already do this the hard way. Every copy-paste round trip to a chatbot costs you flow, formatting and focus, dozens of times a day.
Your best prompts keep vanishing. The instruction that nailed it last month is buried in a chat history. Here it becomes a named, versioned tool you keep.
Your drafts are your intellectual property. Most AI tools share your words by default; Markelangelo makes how much you share a deliberate dial, down to nothing at all.
One page, the whole workshop. Review, rewrite, illustrate, proofread and publish without ever leaving the document you are writing.
It is free while in preview, and it updates itself. Get in early, watch it grow release by release, and help shape what it becomes.
Stop shuttling paragraphs into a chatbot and pasting the answers back. Here the AI works on the page you are already writing.
Start from a prompt, not a blank page. Scaffold a business plan, draft a brief, outline a report, then fill it in.
Make the document talk back. "What is my weakest claim?" "Which assertions have no evidence?" "Answer this from the text only."
Improve, humanise, tighten, re-voice, translate or summarise, on a selection or the whole document, applied in place with a diff.
Real screenshots from the current build, no mock-ups. The manual documents all of it, down to the last control.
One law runs through the whole app: select what you want to do something to, then do it. Right-click a passage and a two-column command centre appears, formatting, synonyms, comment lenses, your prompts and images on the left, and a dedicated column of AI actions on the right: Improve, Humanise, Shorten, Expand, Fix grammar, Tone, Translate and Ask AI. The result streams in beside your text with a word-level diff you approve before anything changes.
Your prompts and Document Leaders are the parts; Flows is where you assemble the machine. Wire them into a pipeline on a visual canvas, then run the whole thing over a selection. One idea flows in, and a full article, a LinkedIn post, an X post, a meta description and a French translation come out, the independent branches running in parallel so a fan-out finishes in the time of its slowest step, not the sum. A run dialog shows its working, ticking each step off with the tokens it used, and every result lands in your document under its own heading. Select an idea, run a Flow, keep writing.
A Document Leader binds a prompt, a persona and a model into one named tool. Select a raw idea, click "Article: Storyteller", and a finished, on-voice draft lands beneath it. Fourteen ship ready-made; building your own takes a minute.
Built for dyslexic writers first. One pass lights up every misspelling together, each with a pale green pill of likely fixes. Click a pill to correct, the cross to keep, or clear the lot with Correct all remaining. Entirely on your machine.
Run a devil's-advocate or evidence-check pass and the AI pins each remark to the exact sentence it is about, right in the margin, like comments beside a Word document. Open one and you are in a focused conversation about that single passage; settle it, then commit the winner in a click: replace the line, keep it as a note, or promote the whole discussion into a linked note.
Select a passage and let the AI draw the illustration it describes, in a house style you set once, saved into a local assets folder beside your document. Or have it suggest the prompt first, or bring your own image from a file or URL.
Work does not arrive as markdown. A client sends a .docx, an accountant sends an .xlsx, and until now that meant closing your editor and opening something else. Markelangelo opens both where you already are: a real word processor with a ribbon, a ruler, true pages, comments and tracked changes, and a real spreadsheet with formulas and formatting. The same prompts and personas you built work in all three. Ask for a rewrite in a Word document and it arrives as a tracked change to accept or reject; ask a question of a spreadsheet and the answer lands in the cells. No import step, no library, no proprietary container: a .docx edited here is a .docx everywhere else.
Write with your notes open beside you. Right-click a tab, choose Split Right or Split Down, and a second fully live editor appears: type in either pane, and every AI tool follows the one you are focused on. Drag the divider, swap the panes, or lock the two documents together with synchronised scrolling to compare drafts line by line. Your layout comes back with your session.
Markelangelo Labs is where the boldest ideas ship first, opt-in, clearly marked, honest about what leaves your machine. Already pay for Claude? Labs lets the text features run on your own Pro or Max subscription instead of an API key: download the runtime on demand, sign in through Anthropic’s own flow, and watch the Max badge light up in the status bar. When a run is covered by your plan, the usage meter shows it for exactly what it is, struck through, not billed.
Press record and talk. Your words build into a live transcript, then one tap tidies them into clean prose or condenses them into a faithful overview, ready to drop into your document or copy anywhere. A large microphone glows soft green when the panel is armed and pulses gold while it is listening, so it is never a question whether the mic is live. It stops itself after a quiet spell, and it is candid about the rest: your audio is transcribed by Groq, whose data, acceptable-use and security policies are linked right there in Labs. Transparency, not fine print.
Type two brackets to link any note to any other, click through to open or create, and watch the Backlinks panel surface every note that points back. Plain markdown underneath, so your web travels anywhere.
A command palette puts the entire app under your fingers, and a quick-action bar in the top toolbar carries your own one-click buttons for the moves you make a hundred times a day.
Three hand-tuned themes and a Typography Studio for the face, size and rhythm of your prose, plus focus, typewriter and zen modes when it is time for the world to fall away.
One click sets the stance. Five settings run from hard to soft, and the app is honest about what each one does and does not promise.
On the cloud stances the routing rules are enforced through OpenRouter and its partner providers, under their published policies; on Totally Local there is no network call at all. How much is shared is always your deliberate choice, never a default you did not make.
How sovereignty works
This is not an app with three toggles in a menu. The whole system opens up in a Settings Shop you can keep open while you write:
These are not hypotheticals. Every one ships as a real prompt in your library, run it, edit it, or fork it into your own.
Your three weakest claims, the strongest counter-argument, and what a hostile reader attacks first, so you fix it before they find it.
Every unsupported assertion flagged, with exactly what proof would make it credible.
Answers your question from the text alone, quotes the passage it relied on, and says plainly when the answer is not there.
Paste the requirements and get a Met / Partial / Missing scorecard with a gaps-to-close list.
You wrote it for engineers; the board needs plain English. Recasts tone, jargon and length, with a diff you approve.
Messy meeting notes in; the decisions actually made and an Action · Owner · Due table out, inventing nothing.
The full workshop costs nothing today. A studio edition is on the way for writers who want more power with none of the setup.
The whole workshop, today. Not a demo, not a tier, the full product.
The studio edition, for writers who want the power without the setup.
Illustrative while we finish the details: final tiers, names and prices may change, and no payment is taken anywhere on this site. The preview app is fully functional and free.
Yes. The app is free while in preview, and it is the full product, the editor, the whole AI workshop, every feature in the manual. A paid studio edition is planned for later; the free app is not a demo.
Only when you run an AI request against a provider you chose, and the Privacy & Sovereignty stance decides how far it may go. On Totally Local, requests run against a model on your own computer and nothing crosses the network at all.
No. Bring your own API key (Anthropic, OpenAI or OpenRouter), or run a free local model through Ollama or LM Studio with no key, no account and no bill.
Windows today, with a self-updating installer. macOS is coming soon.
Windows: Windows 10 or 11, 64-bit, around 600 MB of disk space, and 8 GB of RAM recommended. An internet connection is only needed for cloud AI and the update check. Running a local model via Ollama or LM Studio has its own appetite: allow 16 GB of RAM or more for comfortable results. macOS: coming soon.
They are plain markdown files exactly where you put them on your own disk, with images in an assets folder beside each document. There is no cloud copy and no lock-in; the folder is yours to move, sync or back up however you like.
Keys are encrypted with your operating system’s secure storage and kept on your machine only. They are never synced, never logged, and are used solely to talk to the provider you named.
No account, no sign-up, no telemetry. The only request the app makes on its own is the update check, which asks one question: is there a newer version?
Yes, that is the point. Prompts are editable and versioned, personas capture voice, and Document Leaders bind a prompt, a persona and a model into one-click tools. The manual documents every control.
A place to think, draft, interrogate and reshape, with the AI inside the page and your words under your control. The illustrated guide walks through all of it.
