Expert Advisors
Order state machines that do not double-fire on a requote. Every EA ships with a replay test that runs on 30 days of ticks before it touches live.
- prop-firm rule sets
- multi-symbol routing
- no martingale, ever
Four engineers. No account managers. No “institutional grade” adjectives. We build Expert Advisors, FIX gateways and risk tooling for MT5, and we hand you the logs that prove how they behave.
Four things. If your project is a fifth thing, we will name someone who does it better.
Order state machines that do not double-fire on a requote. Every EA ships with a replay test that runs on 30 days of ticks before it touches live.
Sessions to LMAX, PrimeXM, oneZero and CFH. Sequence resets handled, sessions resumed, partial fills reconciled down to the cent.
Exposure, margin, news windows and drawdown limits enforced before submission. Rejects carry reason codes, not a shrug.
Canvas-drawn dashboards, footprint charts and session tools that stay under 2 ms per tick. No repaint, no memory leaks after 72 h.
A Python driver for the MT5 tester: parameter grids in, tidy reports and honest walk-forward out. It will happily tell you a strategy is bad.
Two weeks, fixed fee, written verdict on someone else's codebase. Roughly half the time the answer is “rewrite it”, and we say so.
Every order gets a UUID at the chart. It survives the gateway, the risk gate, the venue and the reconciliation job. If a hop is slow, the log says which one.
// every hop emits a timed record void OnOrderSent(const ulong ticket) { string id = Uuid(); Log(id, "origin", 0.0); Log(id, "gateway", Ticks()); Log(id, "risk", Ticks()); Log(id, "venue", Ticks()); }
Fixed fee, milestone based, no hourly surprises. Scope changes get a new quote before any code moves.
| model | scope | duration | deliverable | from |
|---|---|---|---|---|
| SPRINT | one EA or one indicator | 2–4 weeks | .ex5 + source + replay test | $6,500 |
| EXECUTION | FIX gateway + risk gate | 8–14 weeks | service, runbooks, reconciliation | $38,000 |
| PLATFORM | white-label MT5 stack | 12–16 weeks | server, CRM glue, portal, monitoring | $85,000 |
| RESCUE | audit of an existing build | 2 weeks fixed | written verdict + fixed-cost plan | $3,200 |
| RETAINER | ongoing support, 20 h/mo | monthly | named engineer, 24/5 window | $4,800/mo |
“They sent a build that failed its own replay test and told us to wait a week. That is the moment we stopped shopping around.”
Four steps. The first one is free and usually ends with us telling you the project is smaller than you think.
You send what you have — a paragraph, a spreadsheet, a Pine script. We reply with questions, not a proposal.
Acceptance criteria first: latency, reject rate, drawdown. The test defines done before code exists.
Private repo you can read, weekly dev builds, changelog per release. No demo-day surprises.
Source, build script, replay manifest, runbook. Then we disappear — unless you keep the retainer.
Demo builds are capped at 0.10 lots and never expire. If it breaks, the issue tracker is public — that is the whole idea.
No forms. No “book a slot with our specialist”. One address, four people reading it.