How to build an app like Wise: Cross-border payments architecture explained
What Matters
- -Wise's technical moat is its network of local bank accounts in 80+ countries, enabling local-to-local transfers without SWIFT fees. Building this network takes years, not weeks.
- -FX rate sourcing is a regulated activity. You can't just pull a rate from a public API. You need a licensed FX provider (or your own license) and rate agreements with liquidity providers.
- -KYC and AML are not features you bolt on. They're regulatory requirements that determine who can use your app, which jurisdictions you can serve, and whether you stay operational.
- -The difference between Wise and a typical money transfer app is float management - Wise earns yield on the float between when users deposit and when transfers settle.
- -MVP scope for a 2-3 corridor cross-border payments product: multi-currency wallets, one payment rail (ACH or SEPA), identity verification, and basic FX conversion. Full Wise parity is a 3-5 year build.
Wise was founded on a simple observation: when founders Taavet Hinrikus and Kristo Kaarmann transferred money between Estonia and the UK, they were losing 5-10% to hidden FX markups from their banks.
Their solution was technically elegant but commercially complex: hold local bank accounts in both countries. Money never crosses borders - it stays local. You deposit in one currency, someone else withdraws in another.
Wise processed $118 billion in transfers in 2024 using this model. But the gap between "simple idea" and "working at $118 billion" is where the real engineering lives.
What Wise actually is
Most people think Wise is an app that converts currencies cheaply. That's the product experience. The infrastructure underneath is different.
Wise operates a network of local bank accounts in 80+ countries. When you send $1,000 from the US to Germany, you don't actually send money internationally. You transfer $1,000 into Wise's US bank account. Wise then transfers the equivalent EUR from its German bank account to the recipient.
No international wire. No SWIFT fees. No correspondent bank markup.
This model requires Wise to maintain liquidity in each corridor - they need enough EUR in Germany to cover outbound transfers. Float management and FX hedging are core operations, not backend features.
Building a full Wise clone means replicating this infrastructure. That's a 3-5 year project. This guide focuses on what's realistic for a startup: a working MVP for 2-3 corridors with a clear path to expansion.
The regulatory layer (read this first)
Most fintech product guides start with app features. This one starts with regulation because regulation determines what you can build, who can use it, and whether you survive audit.
US Money Transmitter Licenses (MTL)
To transfer money in the US, you need a Money Transmitter License in each state where you have customers. That's 48 states (Wyoming and Montana have different structures). Each license costs $5K-$50K and takes 6-18 months to obtain.
Most startups don't apply for 48 MTLs on day one. Instead, they partner with a licensed money transmitter:
- Synapse Financial (banking-as-a-service platform)
- Treasury Prime (bank API platform)
- Column Bank (developer-focused bank with API access)
- Stripe (for US-to-foreign currency flows)
Partnering means you operate under their license while building your product. You give up some margin and some control in exchange for speed to market.
EU Licensing
In Europe, you need either an Electronic Money Institution (EMI) license or a Payment Institution (PI) license from an EU regulator. The UK's FCA, Lithuania's Bank of Lithuania, and Malta's MFSA are common choices for tech companies. EMI licenses cost $50K-$200K and take 6-12 months.
Wise itself holds EMI licenses in the EU, US money transmitter licenses in most states, and a principal member of Visa and Mastercard to issue debit cards. This is a multi-year, multi-million dollar regulatory project.
For an MVP, the path is: start with 1-2 corridors, partner with licensed institutions, build toward your own licenses as volume justifies the cost.
KYC and AML requirements
Know Your Customer (KYC) and Anti-Money Laundering (AML) compliance are not optional features. They're regulatory requirements in every jurisdiction where financial services operate.
At minimum, your app needs:
- Identity verification at onboarding (government ID + selfie)
- Sanctions screening against OFAC, EU, UN lists
- Transaction monitoring for suspicious patterns
- SAR (Suspicious Activity Report) filing capability
Vendors: Jumio, Onfido, and Persona handle identity verification. Sardine and Alloy handle AML monitoring. Budget $2-$8 per verification and $0.10-$0.50 per transaction screened.
Core architecture components
Multi-currency wallet system
Each user has balances in multiple currencies. A USD balance, a EUR balance, and a GBP balance are all separate ledger entries. Transfers between currencies create FX conversion records. Inbound and outbound payments debit and credit the appropriate currency balance.
The ledger is the most critical piece of infrastructure in a payment app. It needs to be:
- Immutable: Entries are never deleted, only reversed
- Consistent: Double-entry bookkeeping - every debit has a corresponding credit
- Auditable: Full history of every state change with timestamps
Use PostgreSQL with proper transaction isolation for the ledger. Don't use NoSQL databases for financial records.
FX rate engine
To quote a transfer from USD to EUR, you need a live mid-market exchange rate. Sources:
API providers: XE.com, Open Exchange Rates, and Fixer.io provide live rates via API. Costs range from $20/month (developer tier) to $1,000+/month (enterprise with SLA).
Bank rates: Your banking partners provide rates for actual settlement. These differ from quoted mid-market rates.
Spread management: Your app quotes the mid-market rate to users and earns revenue from the difference between the mid-market rate and the settlement rate, plus explicit fees.
FX rates in regulated jurisdictions must be disclosed clearly. In the US, the Dodd-Frank Act requires prepayment disclosures for international remittances. In the EU, PSD2 has similar requirements. Build disclosure flows from day one.
Payment rails
A payment rail is the infrastructure that actually moves money between bank accounts. Different corridors use different rails:
| Rail | Use case | Settlement time | Cost |
|---|---|---|---|
| ACH (US domestic) | USD transfers within the US | 1-3 business days | $0.20-$1.50/transfer |
| ACH Same-Day | Faster USD domestic | Same day (3 windows) | $0.50-$2.00 |
| SEPA (EU) | EUR within Europe | 1 business day | €0.20-€0.50 |
| SEPA Instant | Real-time EUR | 10 seconds | €0.20-€1.00 |
| SWIFT | International | 1-5 business days | $15-$50+ |
| FPS (UK) | GBP | Near-instant | £0.10-£0.30 |
| IMPS/UPI (India) | INR | Instant | ₹5-₹15 |
For an MVP, start with ACH (US outbound) and one additional rail for your target corridor. Add rails as you expand.
Transfer state machine
Every transfer goes through a lifecycle. Each state needs to be explicit and auditable:
initiated - user confirmed, debit pending
debit_pending - waiting for funds to clear from user's account
debited - funds received into your account
conversion_pending - FX conversion queued
converted - FX rate locked, target currency reserved
credit_pending - payout initiated to recipient bank
credited - recipient received funds
completed - transfer finished, all records settled
Failed states: debit_failed, conversion_failed, credit_failed, each triggering different refund and notification flows.
Build the state machine before building the UI. Every screen in the app maps to a state or transition.
The feature breakdown
MVP (24-40 weeks, $120K-$250K)
- User registration with KYC (Jumio or Onfido)
- Multi-currency wallets (2-3 currencies)
- Live FX rate display with fee transparency
- Transfer flow with prepayment disclosure
- ACH integration (US funding and payouts)
- One additional rail (SEPA or SWIFT)
- Basic AML monitoring (Sardine or Alloy)
- Transaction history and receipts
- Mobile app (iOS + Android) or web app
- Admin panel: transfer monitoring, KYC review queue, compliance dashboard
V1 (weeks 40-60, +$150K-$250K)
- 5-8 corridors
- Multiple funding methods (bank transfer + debit card)
- Faster payment options (same-day, instant)
- Recipient management (save frequent payees)
- Business account tier (higher limits, bulk payments)
- API access for business users
- Enhanced AML with behavioral scoring
- Rate alerts
V2 and beyond
- Multi-rail routing optimization (choosing cheapest path per transfer)
- Debit card issuance (Visa/Mastercard principal membership or card issuing partner)
- Float yield management
- 20+ corridors
- Direct bank partnerships for better rates
- Business integrations (Xero, QuickBooks, Salesforce)
Tech stack
| Layer | Technology |
|---|---|
| Mobile | React Native or Flutter |
| Web | Next.js |
| Backend | Node.js + TypeScript or Go |
| Ledger database | PostgreSQL (strict ACID compliance) |
| Cache | Redis |
| Queue | RabbitMQ or AWS SQS |
| KYC | Onfido or Jumio |
| AML monitoring | Sardine or Alloy |
| Payments (US) | Synapse, Column, or Stripe Treasury |
| Payments (EU) | Banking partner with SEPA access |
| FX rates | Open Exchange Rates or XE |
| Hosting | AWS (multi-region for data residency) |
What Wise has that you won't have on day one
Being honest about this saves founders from unrealistic roadmaps:
Local bank accounts in 80+ countries: Wise spent 10+ years and hundreds of millions building this network. Your MVP won't have it. You'll use SWIFT for international corridors, which means higher fees and slower settlement than Wise. That's fine - be transparent about it with users.
Principal Visa/Mastercard membership: Wise issues its own debit cards because it's a Visa/Mastercard principal member. Getting this membership takes 1-2 years and requires significant volume commitments. For your MVP, use Marqeta, Galileo, or Stripe Issuing.
$12B in customer float: Wise earns significant interest on the balance sitting in customer accounts. You won't have this at launch. Your revenue model will rely on transfer fees until you reach scale.
Automated compliance in 40+ jurisdictions: Wise has a 400+ person compliance team. Your MVP needs automated tools and a compliance consultant. Hire compliance counsel before you go live, not after a regulatory inquiry.
The corridor to start with
The US-to-India corridor is one of the world's largest remittance flows ($20B+ annually). It's also served by dozens of competitors. High volume, hard to differentiate.
Better starting corridors for an MVP:
- US to specific Latin American market (Mexico, Colombia, Brazil) where you have local knowledge and can build a community
- EU to a specific emerging market where existing services are slow or expensive
- Niche B2B corridor (e.g., US freelancers paying international contractors) where you understand the specific compliance requirements
Pick a corridor where you understand the customer problem and the regulatory environment, not just the volume.
Building a fintech or payments product? Talk to us. We've shipped payment systems and compliance-regulated apps across 100+ products. One call to scope your build.
Related Articles
How to build a trading platform
Read articleAI agents for fintech
Read articleAI in fintech
Read articleFurther Reading
Related posts

How to build an app like Grab: Super app architecture for Southeast Asia
Grab started as a taxi app in Malaysia and became Southeast Asia's most-used app by adding food, payments, deliveries, and financial services in one platform. Here's how the super app architecture works - and how to build one without starting with all of it.

How to build an app like Lyft: Ride-hailing architecture and real engineering challenges
Lyft is not just an Uber clone. It has a distinct pricing model, driver loyalty program, and US-focused regulatory footprint. Here's what makes a ride-hailing app hard to build - and where most teams get the architecture wrong.

How to build a grocery delivery app in 2026: Architecture, features, and what actually ships
You're not building one app. You're building five - and the shopper workflow is the one nobody plans for. Here's the architecture that actually ships a grocery delivery platform to production.