From whiteboard to production

Photograph your whiteboard. Ship the architecture.

A photo of a sketched diagram in, four production artifacts out — rendered diagram, component graph, Terraform skeleton, and ADR draft — in under a minute.

4 artifacts per photo · Mermaid + Terraform + ADR export · EN/FR

Sketch input
Generated blueprint
api-gwauth-svcusers-dbworker
diagram.mmdmain.tfADR-001.md
app.sketchstack.online
Extracted components
  • api-gatewaygateway
  • auth-svcservice
  • users-db (PostgreSQL)db
  • eventsservice
  • queuequeue
  • workerworker

Works with

Whiteboard photoNotebook sketchExcalidraw exportMermaidTerraform (AWS)ADR MarkdownPNG / SVG exportWhiteboard photoNotebook sketchExcalidraw exportMermaidTerraform (AWS)ADR MarkdownPNG / SVG export

The whiteboard tax

Great architecture sessions rarely become shipped specs.

Hours lost

Every design session gets manually redrawn, translated to IaC, and written up — or quietly skipped.

Knowledge evaporates

Architecture lives in one engineer's head and a blurry photo lost in a chat thread.

Docs debt compounds

Undocumented systems get harder to onboard onto every month, until nobody wants to touch them.

What you get back

One photo. Four artifacts you can actually ship.

Everything below is generated from a single hand-drawn sketch — nothing hand-typed.

ComponentTypeProvider hint
api-gatewayGatewayAWS API Gateway
auth-svcServiceAWS ECS Fargate
users-dbDatabaseRDS PostgreSQL
eventsServiceAWS ECS Fargate
queueQueueAWS SQS
workerWorkerAWS ECS Fargate

How it works

Three steps, under a minute.

01step

Snap

Upload any photo of a sketched architecture — whiteboard, notebook, napkin, tablet.

Drop a photo here or click to upload

JPG · PNG · HEIC · up to 10 MB

02step

Extract

AI vision reads boxes, arrows and handwritten labels into a structured component graph.

{
  "components": [
    { "id": "api-gateway", "type": "gateway" },
    { "id": "auth-svc",    "type": "service" },
    { "id": "users-db",    "type": "database" }
  ],
  "edges": [
    ["api-gateway", "auth-svc"],
    ["auth-svc",    "users-db"]
  ]
}
03step

Ship

Get the diagram, Terraform skeleton and ADR — copy, export or push to your repo.

diagram.mmdmain.tfADR-001.md

Your sketch

api gwauthusers dbworker

Generated blueprint

api-gatewayauth-svcusers-dbworker

Try it on a sample sketch — no signup.

Open the live demo

Built for messy reality

Designed for how architecture actually happens.

usr dbusers-db

Handwriting-tolerant vision

Reads rough labels like 'usr db' or 'LB' and maps them to real components.

diagramgraphmain.tfADR

Four artifacts, one photo

Rendered diagram, component graph, Terraform skeleton and ADR draft in a single pass.

spec.v1spec.v2

A spec that stays alive

Re-upload the updated sketch and regenerate the whole spec — no manual diffing.

4

Generated artifacts

2

Export formats (Mermaid, Terraform)

EN + FR

Interface languages

< 60s

From photo to spec (typical)

Why we built this

We kept losing our best architecture sessions to blurry whiteboard photos in Slack threads. Somebody would re-draw them a week later, half wrong, and the original intent was gone. Sketchstack is the tool we wanted: photograph the board, get a real spec back.

We're building in the open — early users get direct access to us. Tell us what breaks and what should exist next.

— The Sketchstack team

Pricing

Start free. Upgrade when the team joins in.

Free

$0/mo

For solo engineers trying it out.

  • 5 diagrams per month
  • Mermaid export
Start free
Most popular

Pro

$29/mo

For architects who ship weekly.

  • Unlimited diagrams
  • Terraform + ADR export
Start Pro trial

Team

$99/mo

For teams sharing a spec library.

  • Shared workspace
  • Version history
Start Team trial

FAQ

Honest answers to the first four questions.

Does it work on messy handwriting?
Yes — the vision model is trained on rough whiteboard sketches. It handles unclear labels by proposing a best guess you can accept or rename before generating artifacts.
Which cloud does the Terraform target?
AWS first. GCP and Azure are on the roadmap. We'd rather ship one provider well than three half-working ones.
Is my diagram data private?
Your uploads are yours. We don't train shared models on your diagrams, and you can delete any project along with its source photo at any time.
Can I edit the result?
Yes. Rename components, retype provider hints, or tweak the Mermaid and Terraform directly — then re-export.

Your next design session deserves a spec.