Showing real prices · what it takes
Showing real prices
is a permissions problem, not a coding one.

Right now every number on the portfolio screen is made up. Replacing them with real ones is easy. Getting permission to show those prices to customers is the hard part — and it decides the cost, the timeline and the design.

THE GOOD NEWS

Abu Dhabi got cheap

The exchange now sells prices directly — about $14 a month.

THE BAD NEWS

Dubai did not

No one sells Dubai prices cheaply. The going rate is around $1,100 a month.

THE TRAP

Cheap plans ban customers

Most low prices are "for your own personal use". Showing them in an app is a separate, pricier deal.

COVERS: UAE · US · crypto · gold · funds PLAN: 4 stages · first one is cheap
Where we are today

Every number on the portfolio screen is invented.

02 / 17
0

Price sources connected

The app has never fetched a real share price. The only outside service it talks to is the AI assistant.

3

Screens showing fake totals

The same invented figure — AED 492,000 — is typed into three different places in the code.

0

Places to store what you own

There is nowhere in the database to record that a customer owns 10 shares of anything.

! Worth saying out loud: we deliberately parked this work earlier to focus the beta on credit cards. Picking it back up is a choice to make on purpose — it isn't a bug someone forgot to fix.
The deciding factor

The UAE is what makes this expensive.

American shares, crypto and gold are cheap and easy to buy prices for. The two local exchanges are not.

03 / 17
Abu Dhabi

The exchange started selling prices directly in August 2026 — about $14 a month, updating live. That is roughly a hundred times cheaper than every other route we found.

Dubai

Nothing comparable exists. You have to go through a big financial data company, and the cheapest quote we found is $1,099 a month for yesterday's closing prices — not live ones.

The usual suppliers

We checked their own data files rather than their marketing. Abu Dhabi simply isn't in them — no ADNOC, no FAB, no Emirates NBD. That is most of the local market missing.

i Both exchanges show 15-minute-old prices free on their own websites — but neither offers a proper way for software to read them.
The survey

We checked thirteen suppliers. Eight are worth showing.

04 / 17
SupplierAbu DhabiDubaiUS sharesCost to show customersVerdict
Abu Dhabi exchangeLive~$14 / moThe only realistic option
Twelve DataYesterday's closeYesterday's closeLive$499 – $1,099 / moOnly one covering both
CoinGecko$35 / moCrypto. Permission is clear
MetalpriceAPI$12 – $32 / moGold. Permission is clear
MarketstackNothingYesterday onlyPartial$9.99 / moNo Abu Dhabi. Terms unclear
FMPNothing20 min old, top planLiveAsk for a quoteEvery listed plan bans apps
MassiveLive~$2,499 / moOnly legal on the top plan
Finnhub · Alpha VantageYesNot soldPersonal use only, every plan

“Cost to show customers” is what we'd actually pay to display prices in the app — not the advertised developer price, which is usually much lower and comes with a ban on customers seeing it.

The trap

The advertised price is not our price.

Nearly every cheap plan says: you may look at this data yourself, but you may not show it to anyone else. Two suppliers spell it out.

05 / 17
Massive

You may not use the Market Data to build an application intended for use by end users other than you.

Their $29 plan behind our app would break that contract outright. The version that permits customers costs about $2,499 a month.

Financial Modeling Prep

Prohibited from showcasing FMP Data on … applications designed for utilization by multiple individuals, irrespective of whether such usage is complimentary or paid.

Every plan they publish — free through $139 a month — is marked “Individual”. Paying more does not fix it. You have to phone them.

! Free Yahoo Finance data is not an option either. Yahoo's terms specifically forbid building an app from it, and the unofficial tool everyone uses says in its own manual that it is for personal study only.
Cost

Everything is cheap except Dubai.

06 / 17
Gold pricesMetalpriceAPI
$12
Abu Dhabi sharesdirect from the exchange
$14
Crypto pricesCoinGecko
$35
US sharesTwelve Data
$499
Dubai sharesyesterday's close only
$1,099
US shares, livethe fully-licensed version
$2,499
$80–120
a month, all in
Gold, crypto, Abu Dhabi and a US source — as long as we accept that Dubai prices are a day old.
If Dubai must be live

The bill jumps past $1,000 a month, and there is no clever way around it. Nobody sells live Dubai prices at a small-business price.

i The first three bars, added together, cost less than one percent of the last one.
Recommendation

One supplier for each kind of thing.

No single company sells everything we need at a price we can legally use. So we buy each piece from whoever does it best.

07 / 17
What we needWho fromCost / monthNote
Abu Dhabi sharesThe exchange itselffree – $14Check their terms before signing up
Dubai sharesTwelve Data, or chase a free delayed feed$1,099 or possibly freeThere is no cheap live option today
US sharesTwelve Data$29 – $499Easiest of all of them to work with
CryptoCoinGecko$35Cheapest with clear permission to display
GoldMetalpriceAPI$12 – $32Permission to display is written down
Fund pricesTiingo$30 – $50Covers 60,000 funds
Order matters. Start with US shares, crypto and gold — cheap, clearly permitted, and they let us build and test the whole system. Add Abu Dhabi once its terms are confirmed. Decide on Dubai last.
How it will work

Prices come through our own server. Never direct.

If the app asked suppliers directly, our paid account keys would be sitting inside every customer's phone — visible to anyone who looked. They also require a single controlled point of delivery.

08 / 17
Customer's app holds no account keys “what's it worth?” Our server Keeps the paid keys never sent to the app Remembers for 60 seconds so we don't ask over and over only if forgotten US shares Twelve Data Crypto CoinGecko Gold MetalpriceAPI Abu Dhabi shares the exchange · later stage each speaks a different language — we translate them into one
i The 60-second memory is not a nicety. Suppliers cap how often we may ask — the cheapest plan allows only eight questions a minute. Without it, a handful of customers would exhaust our quota.
Step by step

What happens when someone opens the screen.

09 / 17
next time, it's already known Opens the screen taps “Portfolio” The app asks “what are these worth right now?” Already know it? YES NO Answer straight away nobody is contacted costs nothing, instant Ask the supplier then remember the answer for the next 60 seconds …and the loop starts again — this time it's already known
i Almost every visit takes the top path. That single fork is what keeps the bill at $50 a month instead of thousands — and it is why a hundred customers cost us barely more than one.
The plan

Four stages. The cheap one comes first.

Each stage is useful on its own, and each one is safe to stop after.

10 / 17
Stage 1

Prices on screen

Live US shares, crypto and gold. No database changes at all. Proves the whole chain works, end to end.

~$50/mo · nothing to undo
Stage 2

Your own holdings

Customers type in what they own. The app works out what it's worth. All the fake numbers get deleted.

database changes · the big one
Stage 3

UAE markets

Abu Dhabi added once its terms are confirmed. Dubai only if you decide it's worth the money.

waiting on two answers
Stage 4

Polish

Everything in dirhams, clear labels on older prices, and the AI assistant able to see the portfolio.

cuts across the others
Stage 1 deliberately comes first because it is cheap, clearly permitted, and throws nothing away. If we stopped there, customers would still get a working price watchlist.
Sequence

What waits on what.

11 / 17
week 1 week 2 week 3 week 4 Safety fix first half a day Stage 1 · prices build · test · ship Stage 2 · holdings database · forms · delete the fakes Stage 3 · UAE blocked until the answers arrive Stage 4 · polish currency · freshness labels · AI assistant
! Only Stage 3 is blocked, and not by engineering — by two questions a person needs to go and answer. Stages 1 and 2 can start tomorrow.
Stage 1 · prices on screen

Cheap, reversible, and it proves everything.

12 / 17
What gets built
  • A translator for each supplier, so the rest of the app sees one consistent kind of "price"
  • The 60-second memory that keeps us inside supplier limits
  • A cap on how often one customer can ask, so nobody can run up the bill
  • One address the app can ask for prices
  • A simple watchlist screen to prove it works
How we'll know it worked
  • Ask for two share prices — both come back current
  • Ask again straight away — the supplier is not contacted a second time
  • Ask far too often — we get a polite refusal, not a bigger invoice
  • Search the built app for our paid account key — it isn't there
  • Prices appear on screen, with older ones clearly marked
i Nothing in this stage touches the database, so there is nothing to migrate and nothing to undo. If the answer turns out to be no, we stop and have lost a few days and about $50.
Stage 2 · your own holdings

Where the invented numbers finally die.

13 / 17
Customer types it in “10 shares of Emaar at 8.20 each” Saved to their record private — only they can ever see it Today's price fetched from whichever supplier covers that market What it's worth quantity × price, added up plus gain or loss since they bought it nothing on this journey is guessed, rounded up, or filled in for effect
Also in this stage

A place in the database to store holdings · a form to add and edit them · the portfolio screen rewritten from scratch · the three fake totals deleted for good.

Two things to budget for

Our current tests write to the real database — that needs fixing before we can test holdings safely. And we have no way to test screens at all, only calculations.

Stage 4 · getting it right

Three rules that keep the numbers honest.

14 / 17
Say how old a price is

A Dubai price from yesterday sitting next to a live American one, added into a single total, is misleading. Every line says how fresh it is — and the total inherits the oldest one.

This matters legally as much as it does for trust.

One currency, converted once

Dirhams, dollars, crypto and gold in one portfolio. We convert on the server, once, and show both the original and the converted figure — never let each screen do its own maths.

The dirham is pegged to the dollar, which covers most of it.

The AI only sees what's allowed

The assistant gets portfolio context the same way it already gets spending: assembled on our server, only with consent, and only as totals — never a list of individual trades.

Same pattern we already use. Nothing new invented.

i These are the rules that separate a real portfolio tool from a convincing-looking one. They cost little to build in now and are painful to retrofit later.
Found along the way

A loaded gun in the database setup.

Unrelated to this project, but anyone doing routine database work could trip it today.

15 / 17

Two live tables are missing from a list they should be on

Our database tool works out what changed by comparing the code to a list. Two tables that exist and hold real data aren't on that list — the audit trail and the spending profiles.

So the tool concludes they were deleted on purpose, and the next routine update it writes will say: delete these two tables. Anyone who runs it without reading it carefully first destroys both.

The fix

Add the two names to the list. It is a two-line change, and it should go in on its own, before any of the work in this deck — not bundled with it.

Before we spend anything

Two questions only a person can answer.

16 / 17
01

Someone opens the Abu Dhabi exchange page

Their website blocks automated visits, so the $14 figure this entire recommendation rests on is unconfirmed. Two things to check while you're there: are we allowed to show their prices to customers, and do they offer the ordinary way of fetching data or something unusual that would take us longer to build?

02

Someone emails LSEG about Dubai

There is a suggestion they give away 15-minute-delayed Dubai prices free. We found it in a search result and could not confirm it. If it's true, it removes the single biggest line from the budget — $1,099 a month, gone. One email is worth the attempt.

? One more unsettled: two readings of Marketstack's terms disagreed — one page is silent about showing data to customers, another seems to allow it. Both are recorded in the written brief.
One boundary worth naming
Showing prices is fine.
Telling people what to buy is not.

Displaying a price, and adding up what someone owns, is ordinary product work. The “AI recommendations” panel on the current portfolio screen is a different thing entirely — telling customers what to buy or sell is regulated financial advice. This plan deletes that panel rather than making it real. Bringing it back is a separate decision for someone qualified to make it.

DECIDE

Restart, or stay parked

Confirm this replaces the earlier decision to leave portfolio work alone.

VERIFY

Settle the Abu Dhabi terms

One person, one browser tab, one email. Everything else waits on it.

RULE

Dubai: yesterday or live

A day old and nearly free, or live for over $1,000 a month. No middle option exists.

DATED: August 2026 — supplier terms change DETAIL: Written brief + full build plan
Presenter notes · slide 01
press N to close