Outwin

Works with your AI

Export a diagram ChatGPT & Claude can actually read.

Every other whiteboard hands your AI a flat PNG — pixels it has to guess at. Outwin hands it the whole picture: an open HTML file that spells out every box, arrow, and note in plain text. Free, browser-only, and no account to start.

The difference in one line

A screenshot shows an AI what your diagram looks like. An Outwin export tells it what your system is — so the answers are about your architecture, not its best guess at a picture.

EVERY OTHER WHITEBOARD OUTWIN diagram.png "is that a database…?" OCR guesswork system.html <box> API Gateway <box> Lambda · Orders <arrow→> writes to <box> DynamoDB <note> retry w/ backoff read as text · no guessing ChatGPT "got the whole flow" Claude "reads every box" Same diagram. One the model squints at, one it simply reads.
Left: a flat screenshot the model has to OCR and guess at. Right: an Outwin .html export whose readable structure drops straight into ChatGPT & Claude.

How it works

  1. 1

    Sketch your system

    Open the canvas and press / to drop boxes, arrows, and notes — plus real AWS, Azure, and GCP service icons and 660+ brand logos. Desktop-first, snappy, and it works offline after the first load.

  2. 2

    Export as HTML

    Hit export and name your file. You get one plain .html file that carries a machine-readable description of every box, arrow, and note — and a PNG too, if you just want a picture for the slide deck.

  3. 3

    Drag it into ChatGPT or Claude

    Attach the .html in any chat and ask away. No converting, no plugins, no screenshots — the model reads the structure and answers about your system.

What you can ask once it's attached

Generate the IaC

"Write the Terraform for this architecture." The model sees every service and exactly how they connect — not a fuzzy picture of them.

Review the design

"Where are the single points of failure? What would you change before this goes to production?" Architecture feedback grounded in the real topology.

Write the docs

"Turn this into a README that explains the data flow end to end." Boxes and arrows become prose, in order, with nothing missed.

Explain it to a new engineer

"Walk a new hire through how a request travels through this system." Onboarding narration, straight from the diagram.

Convert to Mermaid

"Give me this as Mermaid I can paste into the wiki." Because the structure is explicit, the diagram-as-code comes back clean.

Run a security pass

"Flag exposed endpoints, missing auth, and unencrypted hops." The model reasons over the actual connections, not a screenshot.

Why HTML, not an image?

An Outwin export is plain, open markup — not a locked, proprietary file only one app can open. Because it's readable text, the model parses your boxes, links, and notes directly instead of doing OCR on a picture. Four things fall out of that choice:

Readable text, not pixels

The label "DynamoDB" is the word DynamoDB in the file — not a smudge of pixels a model has to recognise. Zero guessing.

Future-proof & portable

It's just HTML — it opens in any browser, on any machine, years from now. No vendor lock-in, no special viewer to install.

Round-trips back into Outwin

The same export reopens in Outwin with every box, arrow, and note intact — so the AI's edits and your diagram stay in sync.

No plugins, no converting

Drag the file straight into ChatGPT or Claude. Nothing to install, no format to translate, no upload pipeline to babysit.

Private by default. Outwin runs entirely in your browser — your diagrams stay on your machine until you hand the file to an AI. No account, no cloud sync, no telemetry watching you draw.

Cloud architecture diagrams Keyboard shortcuts FAQ About Outwin

Hand your AI the whole picture.

Sketch a system, export the HTML, and drop it into ChatGPT or Claude in under two minutes. Free, no account, all in your browser.

Open the canvas