sumsub/agent-skills

sumsub-integrate-travel-rule

End-to-end recipe for integrating Sumsub Travel Rule — deciding which of the three exchange flows applies, identifying the counterparty VASP, submitting the transfer, reading the exchange status, finalising or cancelling it, answering requests other VASPs s…

Quelltext ansehen
Originales Skill-Dokument

Aus dem Quell-Repository gerendert; Überschriften, Beispiele, Code, Tabellen, Links und Bilder bleiben erhalten.

Sumsub — Travel Rule integration

Travel Rule is an exchange of participant data between two VASPs about one transfer. Sumsub is the hub: you talk to Sumsub, Sumsub talks to the counterparty over whichever protocol you both speak.

The integration is small. What makes it fail in production is that most integrations build one half of it — they start exchanges but never finalise them, or they answer the first question and not the second, and nothing in the API complains. This recipe is organised around those omissions.

⚠️ Sandbox tokens only

Do not accept or use a production App Token here. Travel Rule acts on real transfers and real counterparty VASPs, and a test exchange sent to a live counterparty puts a real request in a real compliance queue. Insist on a sandbox pair from <https://cockpit.sumsub.com/checkus/home?sbx=true> — Connect Sumsub to your AI agent -> Build & configure -> Generate token. Token + secret are revealed once; copy both before closing the dialog.

Deeper auth mechanics: `sumsub-api-auth`.

The three flows

Everything starts with one question: who creates the exchange, and has the transfer already settled on-chain?

                          who creates the exchange?
                  ┌────────────────────┬────────────────────┐
                  │        you         │   the counterparty │
   ┌──────────────┼────────────────────┼────────────────────┤
   │ not settled  │  ① before          │                    │
   │ yet          │    settlement      │   ③ you answer     │
   ├──────────────┼────────────────────┤     their request  │
   │ already      │  ② after           │                    │
   │ settled      │    settlement      │                    │
   └──────────────┴────────────────────┴────────────────────┘
① before settlement② after settlement③ answering
Typical caseYour user is withdrawingA deposit arrived with no dataAnything involving your user
info.paymentTxnIdemptyrequiredset by them
Blockingyes — the withdrawal waitsno — funds already movedtheir transfer waits on you
End statefinished after you link the hashfinished, automaticallyfinished, automatically

📘 Do not key your integration off `info.direction`. It says which way the money moves for the account you are looking at, and it is flipped between the two sides of the same exchange. Who created the exchange is the distinction that changes your code; direction is not.

⚠️ Flow ③ is not optional. Every VASP that starts exchanges also receives them, and an unanswered request means your user's transfer is delayed or refused by the counterparty's rules. Integrations that ship only ①/② are the single biggest source of Travel Rule failures.

Stage 0 — Preconditions

Check these before writing any code; each produces a confusing failure later.

  1. Travel Rule entitlement on the tenant — verify with

`sumsub-check-permissions`.

  1. Your organisation is linked to a VASP. Without it, nothing starts and

the error reads Your organization is not yet linked to a VASP. Please contact your Customer Success Manager. That is not a permissions problem and not retryable — relay it and stop. Linking is done by a Customer Success Manager, not over the API.

  1. A Travel Rule rule bundle is installed and active. The bundle decides

what each exchange outcome means for the transaction's review answer — the platform has no default opinion. See `sumsub-create-kyt-rules`.

  1. Travel Rule settings are configured: confirmation timeout, ownership

validation mode, participant data settings.

  1. Webhooks are subscribed and reaching you — see

`sumsub-manage-webhooks`. The events that matter here: applicantKytTxnCreated, applicantKytTxnApproved, applicantKytTxnRejected, applicantKytOnHold, applicantKytTxnAwaitingUser, applicantKytTxnDataChanged.

  1. The Wallet Address Book has your addresses — both the ones you receive

on and the ones you send from. See `sumsub-manage-wallet-address-book`.

Stage 1 — Identify the counterparty VASP

This single step moves the success rate more than anything else in the integration, and it is the one most often skipped.

Option A — let the user pick (recommended)

Add a destination selector to the withdrawal screen and pass the chosen id as counterparty.institutionInfo.internalId. Attribution is then skipped entirely and the request goes straight to the counterparty.

bash
# search the directory as the user types
GET /resources/vasps/-?q=binance&limit=20
  • Search on q — people know brand names, not legal entities.
  • If they pick a group rather than one of its regional entities, that is fine:

Sumsub resolves the group down to the entities underneath it that can actually receive the request. Do not force a choice between subsidiaries.

  • Hidden VASPs are excluded by default; test VASPs only exist in Sandbox, so

no production filtering is needed.

Single VASP by id: GET /resources/api/vasps/{id}/one.

Option B — send only the address

Sumsub attributes it against the wallet address databases, then asks the Travel Rule protocols, then falls back to blockchain analytics providers. This is the right fallback when the user cannot name their destination, but a significant share of addresses cannot be attributed at all.

The two are not exclusive: offer the picker, allow proceeding without it.

Stage 2 — Submit the transfer

Build and post the transaction with `sumsub-create-transaction` using type: travelRule. Only two fields distinguish the flows:

before settlementafter settlement
info.paymentTxnIdomitthe on-chain hash
counterparty.institutionInfo.internalIdVASP id from Stage 1, when knownsame

📘 In flow ② check first that an exchange does not already exist for this transfer — a duplicate puts two requests in the counterparty's queue. Query GET /resources/kyt/txns/query/- walking the deduplication ladder — hash, then address + asset, then amount + time window — and create the exchange only when nothing comes back. Endpoint syntax, the ladder and helper scripts: `references/find-transactions.md`.

🚧 Sending a hash in flow ① turns it into an after-settlement exchange — the counterparty is asked to authorise a transfer that already happened. Omitting it in flow ② does the reverse. This one field decides the semantics.

Currencies and chains

Two currency facts change Travel Rule behaviour, not just precision.

A symbol is not an asset. USDT and USDC each name more than a dozen different assets in Sumsub's catalogue, and over a thousand symbols exist on more than one chain. Send currencyCode without cryptoParams.cryptoChain for one of those and the lookup matches nothing at all, so the asset is unresolved and no conversion happens.

Thresholds are compared against `amountInDefaultCurrency`. If that field is absent, the Travel Rule threshold check is skipped entirely and the flow runs regardless of amount — the same for the unhosted-wallet threshold. The value is also frozen when the transaction is created, so a bad conversion needs a backfill, not a re-read. Send amountInDefaultCurrency and defaultCurrencyCode explicitly whenever you know them.

📘 The counterparty's protocol has its own asset vocabulary, and only a minority of catalogue entries carry a mapping for any given protocol. An exotic token with no mapping cannot be expressed in the outgoing message and the exchange ends at notEnoughCounterpartyData before delivery. Check vocabulary coverage before debugging the payload.

Resolving symbols, chains and aliases: `sumsub-resolve-currency`.

Stage 3 — Read the outcome

Sumsub assigns a Travel Rule status, then your rules assign a review answer. They are different things and readers conflate them constantly: a transaction can be approved by your rules while its exchange sits at expired. Gate your withdrawal on the review answer, and record the exchange status for audit.

Read the transaction with ${CLAUDE_SKILL_DIR}/scripts/get_transaction.sh <txnId> (a signed curl — use the scripts rather than hand-rolling HTTP so signing and encoding stay right). Immediately after submission the exchange either waits or is already final:

OutcomeMeaning
awaitingCounterpartyDelivered; waiting on their answer
counterpartyVaspNotFoundNobody to ask — treated as an unhosted wallet. Usually the largest single outcome by volume; see `references/unhosted-wallets.md`
counterpartyVaspNotReachableIdentified, but no shared protocol
notEnoughCounterpartyDataYour payload failed protocol validation — not their decline
notApplicableYour configuration skipped it

Full status list, transitions and which ones are final: `references/statuses.md`.

Stage 4 — Finalise, or cancel

Only flow ① needs code here. Flows ② and ③ reach finished on their own.

bash
# after you broadcast
${CLAUDE_SKILL_DIR}/scripts/finalize_txn.sh <txnId> <on-chain hash>
# → PATCH /resources/kyt/txns/{id}/data/info   { "paymentTxnId": "<on-chain hash>" }

# if the user abandoned the withdrawal instead
${CLAUDE_SKILL_DIR}/scripts/cancel_txn.sh <txnId>
# → POST /resources/api/tr/{id}/cancel

🚧 This is the most-skipped call in the whole integration. Without it the exchange stops at completed, the counterparty can never reconcile the data they hold against anything on-chain, and neither side has a complete record for audit. Send the hash as soon as you broadcast.

Cancelling matters too: an abandoned exchange left open occupies the counterparty's queue until it expires and records you as unresponsive.

Stage 5 — Answer incoming requests

A request arrives as applicantKytTxnCreated on a transaction where your user is `data.applicant` and the wallet to confirm is data.applicant.paymentMethod.accountId. Fetch it with ${CLAUDE_SKILL_DIR}/scripts/get_transaction.sh <txnId>.

Answering is two responses, in order. The transaction tells you which it still needs:

Flagtruefalse
needMaskingsend response onealready answered — skip
travelRuleInfo.needApplicantOwnershipConfirmationsend response twoalready answered — skip
bash
# response one — is this wallet ours?
${CLAUDE_SKILL_DIR}/scripts/confirm_ownership.sh <txnId> confirmed   # or unconfirmed
# → POST /resources/kyt/txns/{id}/ownership/confirmed

# response two — whose is it?
${CLAUDE_SKILL_DIR}/scripts/attach_applicant.sh <txnId> <applicantId>
# → POST /resources/kyt/txns/{id}/travelRuleOwnership  { "applicantId": "..." }
# an { "applicantParticipant": { ... } } body instead: attach_applicant.sh <txnId> - < body.json

🚧 Order is load-bearing, and getting it wrong fails silently. Response two on its own returns 200 — but while needMasking is true the exchange is pinned at onHold no matter what data you send, the counterparty never receives an answer, and the request expires. Confirming the address is also what unmasks the counterparty's data, so until response one lands there is nothing to reconcile against.

📘 No waiting between them. Both calls return the updated transaction, so there is nothing to re-read and no webhook to wait for.

If you see your transactions sitting at onHold, you are sending response one and not response two.

How long you have

Seconds, not minutes. The counterparty picks the confirmation timeout in their own settings, and the shortest value on offer is currently 10 seconds — the whole round trip, including webhook delivery and both calls, not your thinking time. It is not shown on the transaction and varies by counterparty, so design for the shortest rather than the typical.

Some protocols do not wait at all: the counterparty's system asks and expects the answer inside the same request, so whatever Sumsub can determine from your stored configuration at that instant is the answer. There is no window and no opportunity to answer by hand.

🚧 That is why automation is not an optimisation here. A manual process cannot cover the traffic, and you cannot tell in advance which kind of request is arriving.

Automating both responses

Automation does not replace the two responses — it pre-answers them, so requests arrive with the flags already false.

Register thisPre-answers
Address in the Wallet Address Bookresponse one
Address as an applicant payment methodboth

Neither works unless the ownership validation mode allows reuse — one setting, chosen in the dashboard. See Confirmation ownership mode and `sumsub-manage-wallet-address-book`.

Stage 6 — Test it in Sandbox

Sandbox provides a test VASP that answers like a real counterparty. Work outward from the success case so that when something breaks you know which change caused it, and rehearse the finalisation step specifically — it is the one integrations omit in production.

Procedure, trigger values and what is not reproducible in Sandbox: `references/sandbox-testing.md`.

Go-live checklist

  • [ ] Organisation linked to a VASP; Travel Rule entitlement active
  • [ ] Rule bundle installed and activated, with a deliberate decision for each exchange outcome
  • [ ] Confirmation timeout set to the value you actually want
  • [ ] Webhook endpoint receiving, and idempotent — one exchange emits several events
  • [ ] Withdrawals: VASP picker wired, or a conscious decision to rely on attribution
  • [ ] Withdrawals: the hash is sent after every broadcast, and cancellations are sent on abandonment
  • [ ] Incoming: both responses implemented, in order, with the skip flags respected
  • [ ] Incoming: answered automatically wherever possible; manual is the exception
  • [ ] Wallet Address Book loaded with deposit and withdrawal addresses
  • [ ] Your system distinguishes open statuses from final ones and closes records on the final ones
  • [ ] A failed webhook delivery raises an alert on your side

See also

aus demselben Repository

Weitere Skills

Alle Skills
sumsub
Community

sumsub-analyze-regulation

Analyze a regulation document (PDF or text) and produce a Sumsub configuration plan — mapping regulatory requirements to Sumsub entities (levels, questionnaires, PoA presets, TM rules, workflows, AML resolution rules). TRIGGER when the user provides a regulation PDF, legal act, or compliance requirement document and wants to know what to configure in Sumsub. Acts as the entry point before invoking sumsub-create-level, sumsub-create-questionnaire, sumsub-create-poa-preset, sumsub-create-workflow, sumsub-create-aml-resolution-rules, and other skills. SKIP for direct entity creation requests (no regulatory context) or Sumsub API calls.

Installationen
1
GitHub Stars
6
Aktualisiert
3. Sept.
sumsub
Community

sumsub-api-auth

Authenticate to the Sumsub API with an App Token + secret key (HMAC-SHA256 request signing). TRIGGER when the user asks to "call / sign / authenticate Sumsub API requests", debugs 401 Unauthorized / signature errors against api.sumsub.com, or needs a working request example with X-App-Token / X-App-Access-Sig / X-App-Access-Ts headers. SKIP only when a more specific skill in this repo (questionnaire/level/workflow/POA-preset/generic) already covers the user's actual task — those skills sign requests the same way and only need this one for auth deep dives.

Installationen
1
GitHub Stars
6
Aktualisiert
3. Sept.
sumsub
Community

sumsub-api-generic

Catch-all fallback for any Sumsub API task that does NOT match a more specific skill (e.g. create-sumsub-level, sumsub-create-questionnaire, sumsub-api-auth). TRIGGER when the user wants to call, inspect, or debug a Sumsub API endpoint not otherwise covered — fetching applicants, listing levels, reviewing AML hits, exporting data, generating SDK tokens, anything against api.sumsub.com. The procedure — locate the right endpoint in the OpenAPI schema, read its request/response shape, build the payload, sign with App Token, and validate. SKIP whenever a narrower Sumsub skill already covers the request.

Installationen
1
GitHub Stars
6
Aktualisiert
3. Sept.
sumsub
Community

sumsub-create-aml-resolution-rules

Create, edit, reorder, delete, and publish Sumsub AML Resolution Rules (the AML Resolution Rule Chain) that auto-review AML screening hits. TRIGGER when the user wants to auto-clear false positives, auto-confirm true positives, carry over previous AML reviews, tag AML hits, or set up / inspect / publish the AML rule chain. SKIP for transaction-monitoring (KYT) rules, workflow routing, or AML check settings on a level (separate skills cover those).

Installationen
1
GitHub Stars
6
Aktualisiert
3. Sept.