1
User:
Claude
Waiting for prompt…
···

Open a file above to begin the demo

v2.74
⚠️
Notice
📦
One more click coming up
Right after this, Chrome will ask if you trust this download. That's normal — .vsdx is a compressed file type, and Chrome shows that check for every compressed file, from every site. It isn't a problem with this file. Just click Save when you see it.
EU Artificial Intelligence Act · Regulation (EU) 2024/1689 · Art. 50(1) · Applicable from 2 Aug 2026
🤖
You're interacting with an AI system
This tool uses Claude (by Anthropic) to read, convert, and modify your process diagrams. The diagram code, objective, and explanation shown after each Apply or file import are generated by AI, not written by a person.
AI output can be wrong. Review any diagram before relying on it for a real business process, especially before using it to make compliance, financial, or operational decisions.
Lo-Fi (Saver) vs. Hi-Fi (Pro)
Two ways to get the same great diagram. Pick based on what matters most right now.
🔒

What happens to your file

Hi-Fi sends your file through exactly as uploaded, nothing changed, nothing left out. Lo-Fi flattens SVG files into a plain picture first, which has a nice side effect: anything hidden inside the file, like a script or an outside link, never makes the trip.

🧩

You always get a real, editable diagram

Either mode hands you back a diagram you can keep working with, never a pile of disconnected shapes. Save it as .drawio, Visio, or BPMN and pick up right where you left off, anywhere. The one thing to know: on a busy, detail-heavy diagram, Lo-Fi is reading from a picture instead of exact text, so it can occasionally misread a small label.

💰

Same quality, smaller bill

Lo-Fi writes a shorter summary and turns SVGs into a picture first, both mean less work behind the scenes, which means it costs less. Perfect for your everyday, routine diagrams.

Reach for Hi-Fi when every word has to be exact. Reach for Lo-Fi to save on the simple, everyday stuff.
Obfuscation

Heads up: in this version, Glossary Ingestion only reads file formats we can interpret locally. No LLM involved, no internet traffic, no Claude token cost. A future version, using a local LLM, will widen this list.

Included (read locally, $0)
  • .mmd, .mermaid (Mermaid source code)
  • .md, .markdown (with a Mermaid code block inside)
  • .bpmn, .drawio, .svg (read locally as XML, no LLM)
  • .vsdx (unzipped locally, no LLM)
Not included
  • Images (.jpg, .png, .webp, .gif, .heic)
  • .pdf
  • .xml, .html
  • Other formats (.docx, .pptx, audio, Photoshop files)
Still work normally in Production, via Open File + Apply.
Obfuscation = smart protection against leaking info to the LLMs
Setting up the Glossary is quick. Open a handful of your usual diagrams, let the system scan them, remove what does not matter, and you are ready. Soon you will be able to point this at a whole folder and let the system work through every diagram and flowchart on its own, so this step gets even lighter for you and your team.

Two modes work together here:
  • Glossary Ingestion ($0): a training pass. Open your diagrams and the system looks for capitalized names, terms, and phrases, the usual shape of a proper noun or product name, and adds them to the list below for you to review. Nothing gets sent to Claude while you're in this mode, so there's no cost.
  • Production Mode: your normal working mode, once your list looks right. The Glossary you built applies quietly in the background to keep those words out of every request.
Here is what Obfuscation actually does behind the scenes.

A process in our system temporarily replaces the names, terms, and phrases in your Glossary, the ones you decided are worth protecting, with codes that mean nothing on their own (e.g. Claude, OpenAI, Grok, Gemini, etc.). For example, "Acme Partner" becomes "C0035" in the Glossary, and "Initiate Project Lunar" becomes "C0203". For you, none of this shows: you always see the full diagram or flowchart with the real names, and that's exactly how it gets saved when you export or "save as" it too.

In other words, on top of the policies these LLMs already publish about not using submitted data for training, the Obfuscation process adds one more layer. It keeps the words in your Glossary from ever traveling across the internet and reaching the LLMs used in the GS4D system.

And here's the best part: when you save your diagrams and flowcharts, the system reconstructs all the words, terms, and jargon that were temporarily replaced. The content stays exactly as it was.

If the computer you're using is shared with other people, once you're done with GS4D you should still clear your browser's cache and stored data.
Works on the diagram text and your Apply requests. It can’t redact a name baked into an uploaded image or PDF. Use a text-based file (.drawio, .svg, .bpmn, .mmd) if that matters for your diagram.
Redaction Glossary
Real termPlaceholder / code
🗑️
Reset Glossary?
This will permanently erase all Glossary pairs on this browser. This can't be undone.
Warm vs. Cold Cache: the free discount hiding in your workflow

Work in a burst, pay less

Run all of today's diagrams back-to-back, inside the same 60-minute window, and every import after the first one comes in cheaper and faster, automatically, no extra step needed.

Kicking off a big batch?

Start with one import to get things warmed up, then queue the rest right behind it. That first one pays a small, one-time setup cost. Everything else you run in the next hour rides the discount.

Don't let it go cold

Wait more than an hour between imports and the discount resets, even on the exact same diagram. If cost matters to you, batch your work instead of trickling it in through the day.

This is only about timing, nothing to do with your Lo-Fi/Hi-Fi choice. Just: how long ago was your last import?