Bob in Dubai pays Sweety in Bangalore. Their banks have no account with each other and don't even share a currency. Watch correspondent banks, a cover payment, and an FX conversion stretch the simple transfer across borders without losing the thread.
Read in the interactive Academy →Dirhams out, rupees in. Bob is in Dubai with dirhams; Sweety is in Bangalore expecting rupees. Their two banks have never dealt with each other — no shared account, no direct line, not even the same currency. In the domestic transfer both banks plugged into one rail and were done in seconds. Here, the rail runs out at the border.
This is the customer transfer again — pain instructs, pacs settles, camt reports — with the hard part of international banking dropped in the middle: the two end banks have no relationship. Everything new here exists to bridge that gap, and you can reason out each piece.
{{think}} Bob's bank in Dubai holds no account with Sweety's bank in Bangalore. A bank can only pay another bank it holds an account with (a nostro/vostro relationship). The domestic rail that connected them last time doesn't reach across the border.
So how does the money get from one to the other? {{reveal}} Through banks they both deal with. Bob's bank and Sweety's bank each hold accounts with larger banks that hold accounts with each other — correspondents. Chain those links and money can travel between any two banks on earth, even ones that have never met:
Bob's bank (Dubai) → a correspondent bridging AED and the international network → Sweety's bank (Bangalore)
Each link is a pair of banks that genuinely hold money for each other. The chain is the route. {{/think}}
{{think}} Now the subtle part. When banks route serially through correspondents, the payment travels as two parallel messages, not one. Why would you separate the payment's information from its funds — and what stops the two from drifting apart? {{reveal}} Because the information needs to reach Sweety's bank directly and fast (so it knows who to credit and why), while the actual money settles between the correspondents that hold accounts for each other. So:
Two messages, one payment. What stops them drifting apart is the shared UETR on both legs. Lose it and the cover can't be matched to the customer payment. {{/think}}
Somewhere along the chain, dirhams become rupees. FX happens at the bank holding both currencies (usually a correspondent, sometimes Sweety's bank), and the message records it explicitly: the instructed amount (what Bob sent, AED), the exchange rate, and the settlement amount (what arrives, INR). Charges get deducted somewhere too, recorded in the charge information so everyone sees who paid the cost of the crossing.
BOB-INV0042, beneficiary Sweety in India.UETR; each correspondent passes the information along.UETR; FX is applied at the bank holding both currencies.BOB-INV0042, so despite the currency change and the chain of strangers she still matches Invoice 0042.Open the pacs.009 cover payment →
The cover leg, field by field: pacs.009 up close →
The shape is identical to the domestic transfer; three things were added to cross the border — a chain of correspondents (the end banks have no direct link), the funds split from the information (pacs.008 + pacs.009 COV, reconciled by a shared UETR), and FX and charges entering the message. And still the thread held: the EndToEndId Bob typed in Dubai surfaced in Sweety's notification in Bangalore, unchanged across borders, currencies, and a chain of banks that had never met. That is the whole point of a globally-shared reference — a payment can be handed between strangers and stay, end to end, one payment.
{{aside:model|The mental model}} Cross-border = the domestic transfer plus a correspondent chain. The end banks have no relationship, so the payment splits: a pacs.008 customer leg (the information) and a pacs.009 COV cover leg (the funds), reconciled by a shared UETR. FX and charges enter the message; the EndToEndId survives the whole crossing. {{/aside}}
{{aside:chair|From the engineer's chair}} Two routing methods exist: serial (one pacs.008 hand-to-hand) and cover (pacs.008 direct + pacs.009 COV behind), and only the shared UETR links the cover to the customer leg — reconcile on it. FX shows up as three fields (instructed amount, rate, settlement amount); charges are recorded so SHAR/DEBT/CRED is auditable. (Our Playground engine transforms the pacs.008 customer leg today; the pacs.009 cover leg is a natural next mapping.) {{/aside}}
{{aside:breaks|Where it breaks}}
{{/aside}}
UETR. Then the cover funds and the customer payment can't be matched — money with no story, straight to an investigation.{{aside:map|The map}} A border on top of the base case:
{{/aside}}
{{aside:ref|Reference card}}
{{/aside}}
pacs.008 (customer info) + pacs.009 COV (funds), reconciled by a shared UETR.EndToEndId survives borders and currencies — still one traceable payment.Inspect the cross-border legs in the Playground →
You can walk a cross-border payment through correspondent banks, explain why intermediaries appear, tell the customer leg (pacs.008) from the cover leg (pacs.009 COV), see where FX turns one currency into another, and follow the shared UETR that reconciles the two legs. You can see that cross-border is the domestic transfer plus a chain, and that the EndToEndId survives the whole crossing, keeping it one payment from Dubai to Bangalore.
Why do cross-border payments often pass through correspondent banks? The sender's and receiver's banks may hold no direct relationship or shared currency accounts
What lets a cross-border payment be tracked across every hop? A globally unique end-to-end reference that every bank preserves