Currently offline

Plan / Prep / Plate

A recipe app that actually shipped. Revenue just never followed.

CookCloud was a full meal-planning platform: search, import, plan, shop, track. The product worked. Getting people to pay for it never did, partly because I kept shipping new features instead of fixing that. So I pulled the plug. Here's the whole thing, unfiltered.

Get notified if it comes back Read the original launch post →
Plan · Prep · Plate

What actually shipped

This isn't a landing-page mockup. It's a working product people used, and everything marked shipped below ran in production.

Search by keyword or by ingredient

Type in what's already in the fridge. Recipes come back ranked by how many of those ingredients you actually have, so it's not just matching a keyword to a title.

shipped

Custom recipe creation and import

Write your own recipes from scratch, or drop in a URL and CookCloud parses the page and structures it automatically.

shipped

Personalized meal planning

Drag recipes onto a weekly calendar. That plan is the source for the grocery list, the nutrition totals, and the prep schedule, so everything downstream stays in sync.

shipped

Automated grocery lists

Every ingredient across a week's meal plan gets deduplicated and quantity-summed into a single shopping list. No manual tallying.

shipped

Ingredient substitutions

Out of buttermilk, or allergic to shellfish? A dedicated substitution engine matched individual components, down to specific ingredients, with real alternatives.

shipped

Nutritional tracking

Calories and macros roll up per recipe and per day, straight from the plan you already built. There's no separate food diary to keep up with.

shipped

Meal Prep Mode

A focused, step-by-step cooking interface, built for a phone propped up on the counter rather than a browser tab you have to keep scrolling.

shipped

Smart recommendations

A recommendations engine got scaffolded to suggest recipes based on what you'd cooked, saved, and searched before. It stalled in early development and never made it to users.

roadmap

Pantry tracking, barcode scanning, delivery hookups

Scoped, prioritized on a public roadmap, voted on by real users. I kept starting new work before finishing what was already sitting there, so none of it got built.

roadmap
Under the Hood

What ran underneath it

The launch post covered the AI coding tools I built it with. This is the part it didn't cover: what actually had to run, all day, every day, for CookCloud to stay up.

Application

  • Django 5.2— core framework, Python 3.13
  • Django Channels + Daphne— async/WebSocket layer
  • Django CMS— blog & marketing pages
  • django-allauth— auth, MFA, social login
  • DRF + drf-spectacular— API + docs

Data & jobs

  • PostgreSQL— primary datastore
  • Redis— cache, broker, sessions
  • Celery + Celery Beat— background & scheduled jobs
  • django-storages + S3— media & image storage

Intelligence

  • OpenAI API— recipe parsing & suggestions
  • Spoonacular API— nutrition & ingredient data
  • AbacusAI— recommendation modeling
  • yt-dlp— recipe import from video

Money & ops

  • Stripe + dj-stripe— subscription billing
  • Mailgun— transactional email
  • Sentry— error tracking
  • GeoIP2 + django-health-check— access & uptime
This is a real SaaS stack, not a script. Postgres, Redis, a Celery worker, a Celery Beat scheduler, and an async web process all had to run continuously, at the same time, just for the app to function. Five always-on processes before a single user loads a page. It's the right architecture for what CookCloud does, and it meant the app needed paying subscribers to make sense financially (more on that below).
The Honest Part

Why the kitchen closed

Revenue never really got off the ground. Five always-on services and three metered APIs need paying subscribers behind them, and I spent my time building instead of finding those subscribers.

Paying subscribers at shutdownzero
Always-on services required to run the appfive
Shipped features (search, import, planning, lists, subs, nutrition, prep mode)seven
Of those, polished enough to convert a visitor into a subscriberfew
Weeks spent marketing what already existednot enough
A real product that never found paying usersthe actual problem

CookCloud never found paying users. That's the plain version, and everything else is a more comfortable way of saying it. A recipe app running on Postgres, Redis, Celery, and metered AI APIs needs subscribers behind it to make sense, and revenue just never got there.

The mistake wasn't the tech. It was where I spent my time. Every time a core feature felt slightly rough, my instinct was to build the next feature instead of polishing the one already in front of me. Search, meal planning, and grocery lists all deserved more attention than they got before I moved on to ingredient substitutions, then started scaffolding a recommendations engine nobody had asked to use yet. Telling people CookCloud existed barely happened at all.

So I took it down rather than let it quietly bleed money every month. The code still works, and the idea still holds up — but a relaunch means doing the boring part first: pick two or three features, make them genuinely great, and actually tell people about it before starting anything new.

Reservation list, not a mailing list

If enough people want it back, I'll bring it back.

One email, no spam. If this list gets big enough to justify the hosting bill, you'll be the first to know CookCloud is cooking again.

Used only to notify you about CookCloud. Unsubscribe anytime. Privacy policy