Analytic Endeavors Design - Copyright 2024-2026 Analytic Endeavors Inc. Unauthorized use prohibited.
Private beta
Diagram your data stack in seconds.
Describe a data flow in a sentence and get a clean, branded diagram you can drop straight into a deck. Generated by the Analytic Endeavors diagram engine, tuned for the Microsoft data stack.
Your description is sent to our diagram generator to build the layout. Keep it high-level. Don't paste credentials, connection strings, or confidential data. The example buttons run entirely in your browser.
Need this built for real?
Production architectures, semantic models, and pipelines, designed and delivered by the Analytic Endeavors team.
Get a link to reload and keep editing this diagram. We'll also send you the occasional data & AI guide (unsubscribe anytime).
Saved. Here's your link
Open it anytime to reload this diagram and keep editing.
Check your email
We sent a one-time sign-in link to . Click it to unlock 10 AI edits per diagram. The link expires in 15 minutes.
How to prompt the diagram engine
Describe your architecture in plain language. The more you name the real systems and how data moves between them, the closer the first draft lands. Here is the vocabulary it understands.
Stages & flow
Name your stages left to right and they become lanes: sources, ingest, lakehouse, serve.
Keep nodes to the real objects. A semantic model is one box, a medallion is three (bronze, silver, gold).
How data moves
Streaming (moving dashes): say mirror with CDC, Auto Loader, Eventstream.
Batch (solid): say nightly ELT, import-mode refresh.
Reference, no copy (dotted, hollow arrow): say Direct Lake, DirectQuery, shortcut, Delta Sharing.
Boundaries & colour
Wrap items in a frame: put bronze, silver, gold in a OneLake boundary or a workspace around the model and reports.
Frames can nest: a OneLake boundary with bronze, silver and gold zones inside.
Colour by layer happens automatically. Say mark the legacy systems as at-risk for a red, retire-it look.
Side-by-side & layout
Put peers on one row: a lakehouse and a warehouse side by side in OneLake, or Salesforce and HubSpot together as sources (up to four per row).
Everything is editable by hand: in Edit by hand a lane stacks rows, each row holds 1-4 cards, and you can set each row's left/center/right alignment and each lane's top/middle/bottom position.
Movement & loops
Direction and life: a fast streaming line from events to bronze, or show the refresh as a slow batch.
Feedback: write predictions back to gold or reverse-ETL to Salesforce draws a loop.
Add an AI box by naming MLflow, a copilot, or Genie.
Descriptions & tooltips
Ask for hover notes: add short descriptions to the key boxes, or explain the bronze, silver and gold layers.
A box with a description shows a small info marker - hover it to read the note.
Markdown works inside a description: **bold**, *italic*, `code` and [links](https://...).
After it generates, click any box to spotlight everything it touches, or open Edit by hand to fine-tune line types, animation, colours, labels and caption position, add side-by-side cards and rows, draw frames (dashed/solid/dotted, with the title where you want it), set lane alignment, and give any box a description that shows as a hover tooltip (with a small info marker on the card). Then export to PNG or SVG.
About this tool, your data & security
A free tool from Analytic Endeavors. Here is what happens under the hood, where your text goes, and how it is kept safe.
How it works
You describe an architecture in plain language. That text goes to our generator, which asks a language model to draft a layout, and returns a structured diagram spec. The diagram itself is drawn in your browser by the Analytic Endeavors diagram engine. The example buttons run entirely in your browser with no network call at all.
Your data & the language model
Your description is processed by Anthropic's Claude (the company behind Claude) through their API to draft the layout.
Per Anthropic's API terms, prompts sent through the API are not used to train their models.
We do not store your prompt. Our server logs only its length (a number), never its content.
Please keep descriptions high level. Do not paste credentials, connection strings, or confidential data - the tool never needs them and never asks for them.
If you use Save & share, we store that one diagram and your contact details so we can generate your reload link. Nothing else is retained.
If you sign in, we keep a one-way hash of your email so we can recognise a returning sign-in (and send the occasional guide if you opt in). We never store your password because there is none - sign-in is a one-time emailed link.
Security
API keys live server-side only. They are never sent to your browser.
The model is scope-locked: it only ever returns a diagram spec. It will not answer questions, follow instructions hidden in a prompt, or reveal how it works. Every response is validated and bounded on our server before it reaches your screen.
Requests are rate-limited and capped per day to keep the free tool fair and abuse-resistant.
This design, the diagram engine, and the generator are proprietary to Analytic Endeavors and carry embedded provenance markers.
Credits
Icons are from Lucide, a free open-source icon library (ISC license). Diagram rendering, animation, and export are the Analytic Endeavors engine.
Paste a diagram spec JSON (the same shape the Copy spec button produces) to render it as a fully editable diagram. No AI call, runs entirely in your browser. Handy for tweaking a saved spec or one drafted elsewhere.
Send feedback
Found a bug, hit a wall, or have an idea? It goes straight onto the team's board with the context attached.
Attachments help us reproduce exactly what you saw. Nothing else about your session is sent.
Thank you - filed.
Your feedback is on the team's board. During the beta it genuinely shapes what gets fixed next.