Forex Delight is the most technically complex project I’ve built to date. A cloud-powered trading ecosystem that connects live MetaTrader 5 accounts to a web platform, processes real-time trade data, and provides analytics dashboards to traders managing their growth.
Building it taught me things that theoretical knowledge doesn’t. This post covers the real challenges, the decisions I made, and what I’d do differently.
The core challenge: real-time financial data
The hardest part of building a trading platform is not the UI. It is the data pipeline.
Trade data in MetaTrader 5 lives inside the terminal — a Windows desktop application running a trading account. Getting that data out of the terminal and into a web platform requires bridging two completely different worlds: a legacy desktop application with its own scripting language (MQL5) and a modern cloud API.
The solution was a custom Expert Advisor (EA) — a program written in MQL5 that runs inside MT5 and pushes trade events to a Cloudflare Workers API endpoint in real time.
// Simplified EA logic — fires on every new trade event
void OnTradeTransaction(const MqlTradeTransaction &trans, ...) {
if (trans.type == TRADE_TRANSACTION_DEAL_ADD) {
string payload = BuildTradePayload(trans);
SendToAPI(payload, "https://api.forex-delight.com/trades");
}
}When a trade opens or closes, the EA fires, builds a JSON payload, and sends it to the Workers endpoint. The response is typically under 500ms — fast enough for real-time tracking without blocking the terminal.
Why Cloudflare Workers for the backend
The choice of Cloudflare Workers for the API layer was deliberate. Trading data is latency-sensitive — delays in recording trades create inconsistencies between what MT5 shows and what the dashboard shows. Users notice immediately.
Workers run at the edge — in the Cloudflare location closest to wherever the EA is sending data from. There are no cold starts. Response times are consistently low. For a UK-based trader, requests hit a London edge node. For a Nigerian trader, Lagos. The infrastructure adapts geographically without configuration.
The database decision
Cloudflare D1 — serverless SQLite at the edge — was the right choice for trade storage. The data model is straightforwardly relational: accounts, trades, performance metrics. SQLite handles this cleanly. D1 integrates directly with Workers, removing the latency of an external database call.
The trade-off: D1 is not suited for high-write workloads or complex analytical queries. For a platform with hundreds of concurrent traders executing automated strategies, a more powerful database (CockroachDB, PlanetScale) would be the next step.
What I’d do differently
Build the data schema more carefully upfront. The trade data model evolved three times during development as I understood the MT5 output format better. Each schema change required migration work that could have been avoided with more careful design at the start.
Add rate limiting from day one. The EA can fire frequently during active trading sessions. Without rate limiting on the Workers endpoint, a misbehaving EA could generate thousands of requests per minute. I added rate limiting late in development — it should have been there from the first deployed endpoint.
Build the monitoring before the features. I added Cloudflare analytics and custom metrics logging partway through development. I should have set this up on the first day. Debugging data pipeline issues without visibility into what the Workers were receiving and rejecting was unnecessarily difficult.
What it proved
Cloudflare Workers is production-ready for real-time data applications. The latency is genuinely low. The integration with D1 is clean. The deployment pipeline (GitHub → Cloudflare Pages/Workers) is fast and reliable.
MQL5 EA development is a niche skill, but it is not as opaque as it appears. The language is C-like and the documentation is comprehensive. The hard part is understanding the MT5 event model — once that clicks, the development is straightforward.
Financial platforms are buildable by a solo engineer with the right architecture decisions. Forex Delight is not a team effort. It is a single engineer choosing the right tools and building systematically.
If you’re working on a fintech platform or a system that needs real-time data integration, let’s talk. I’d rather you learn from my mistakes than repeat them.