ROKFX — Most Brokers Build a Counter. We Built a City.

Every brokerage you've ever used is, architecturally, a counter: you walk up, place your order, and walk away. The relationship is transactional because the building is transactional — there's nowhere to stay.

Now imagine a brokerage built as a city instead. A bank district where your money moves under guard. An exchange plaza where partners run distribution routes. A tower where managed strategies live. And at the center, a place where you live — an avatar that feels your P&L, levels up on your real trades, and competes in the arena on weekends. When there's somewhere to stay, then traders stay.

That's ROKFX, and Rocket Now built it — twice, actually. Once in Postgres. Once in WebGL.

The Client

ROKFX set out to run a retail multi-asset brokerage that refuses to stop at order routing. The commercial thesis has three parts, and each is normally its own product team: a complete client lifecycle (signup, KYC, wire and crypto funding, multi-wallet balances, withdrawals under treasury control, in-app support); a distribution engine (an introducing-broker partner network with referral tracking, automated commission math, payouts, and marketing collateral — because acquisition economics decide whether a brokerage survives); and a differentiated retention layer — Tradetopia, a 3D world where your avatar reacts emotionally to your live P&L, earns points from real trades, unlocks floors of the Trading Tower, copies other traders, and competes for prize pools.

Three product teams' worth of scope, plus the admin and treasury back office that finance teams and regulators require. Conventionally, that's a multi-year roadmap. ROKFX needed something real to show partners and regulators — and needed it at the speed of conviction, not the speed of headcount.

The Challenge: A City's Worth of Specialisms

Look at what this build actually spans: payments and compliance workflows, commission accounting, managed-account fee math, real-time multiplayer, and 3D rendering. When the functional surface crosses that many specialisms, then no small team has them all in-house — and hiring them all means the roadmap dies of coordination before it dies of anything else.

Two constraints sharpened everything. First: money movement demands correctness — deposits, transfers, commissions, PAMM fees, and treasury settlement are all paths where a defect is a financial loss, never a cosmetic bug. Second: five kinds of citizens — clients, partners, PAMM managers, admins, treasury operators — must each see only what their role permits, enforced at the data layer, because a UI check is a suggestion and a database policy is a law.

The Build: One City, Six Districts

Rocket Now ran the engagement as a Claude-assisted engineering program, on a simple piece of judgment: this work is overwhelmingly well-specified but voluminous — dozens of CRUD-plus-policy surfaces, a large relational schema under strict row-level security, and a very large amount of interface code. That's exactly the profile where an AI-assisted loop compounds fastest, provided a human engineer owns specification and verification. So every phase began as a written implementation spec — 28 phase documents plus the partner-system program, committed alongside the code — and Rocket Now engineers retained line-by-line ownership of scope, financial logic, security, and release decisions. Claude was the accelerator, never the approver.

The delivered city, by district:

The residential district — the client platform. Dashboard, trading-account creation with TradeLocker integration, bank-wire deposits with reference matching, crypto deposits with generated addresses and automated on-chain monitoring, withdrawals gated by verification codes, internal and cross-wallet transfers, multi-wallet views, consolidated history, profile and KYC, a notification centre, and support with live chat.

The roads — the partner/IB system. Application and approval workflows where approval atomically grants the role, mints the referral code, and provisions the partner wallet in one database transaction; click-tracked referral links; configurable commission rates with per-trade calculation and an approval queue before anything pays out; payouts, marketing libraries, and exportable reporting.

The tower — PAMM. Strategy creation and discovery, investor onboarding with investment modelling, performance charting, scheduled management and performance fee collection, structured withdrawal requests, and private-partner allocations.

The bank — admin and treasury. A full back office of approval queues, user administration through privileged server-side functions only, audit logging — and money egress deliberately walled into its own district: a dedicated treasury guard and a three-stage pipeline where a withdrawal is initiated, then approved, then executed as separate authenticated steps. When money leaves the city, then it leaves through three locked gates, each with its own key-holder.

The arena — Tradetopia. Delivered across 24 core phases and 4 expansions on isolated routes so it can launch independently: the 3D city with five landmark buildings, the five-floor Trading Tower with elevator progression, a rigged avatar with Ready Player Me integration whose emotion system runs off live P&L, points earned only from real trading, skill trees, quests, a customisation economy, leaderboards, friends, database-backed copy trading linked to real accounts, competitions with entry fees and prize pools, multiplayer presence, replay capture, and share cards. Retention isn't a re-engagement email here — it's a place.

The foundations under all of it. Roles in a dedicated table read through security-definer checks; row-level security with explicit grants on every one of the 84 tables; privileged operations server-side only; verification codes on sensitive actions; typecheck, lint, and database-lint gates on every change.

The totals: 77 routes, 208 components, 87 typed data hooks, 36 edge functions, 84 tables across 73 incremental migrations, ~72,800 lines of application code — live at rokfx.com with a separate preview environment, now moving through a staged rollout: hardening and a line-by-line money-path review, staging UAT across all five roles, an invite-only pilot with deposit caps and daily reconciliation, then general availability, then Tradetopia's public launch.

The Results

Six product domains delivered as one platform — client, partner/IB, PAMM, admin, treasury, and a 3D gamified world — scope that conventionally consumes several teams and a multi-year roadmap
Every money path built for correctness: atomic role-and-wallet provisioning, commission approval queues, three-stage treasury settlement, and human line-by-line review of all financial logic
Security as law, not suggestion: RLS with explicit grants on all 84 tables, five-tier route guarding validated server-side, and zero privileged operations reachable from the browser
A written engineering record: 28+ phase specifications that double as the platform's living documentation
Deployed to production with a disciplined five-stage rollout underway

"[Quote from ROKFX leadership — pending sign-off.]" — [Name, Title], ROKFX

Why It Worked: One Right Step, Then a City

The Rocket Now principle, one more time, because every engagement proves it again: a single step in the right direction is worth exponentially more than many steps in the wrong one. For ROKFX, the right step was taken before any code existed — specification before implementation. Every phase written down first meant every large AI-generated change was reviewable instead of opaque; reviewable changes meant velocity without fear; and velocity without fear is how a small team ships a city. When the spec leads and the human owns the money paths, then AI-assisted delivery stops being a gamble and becomes a multiplier — and the roadmap that read as "multi-year" reads as "delivered."

Carrying a roadmap that spans more specialisms than your team? That's a city-building problem. Bring us the map. → RocketNow.com

Editor's note — before publishing: