Skip to content
Velloreth
Financial Technology 11 min read

Understanding Banking APIs and Open Finance

Banking APIs turned bank data into something third parties can use with consent. How open banking and open finance work, where the standards are, and what actually builds on top.

By

Open banking starts from an uncomfortable observation: the institution that knows the most about your financial life — your bank — has historically been the institution least willing to let anything useful happen with that information. Moving money out required branch visits or clunky net banking. Sharing income proof for a loan meant exporting a PDF, getting it stamped, and handing it to a stranger. Building anything on top of bank data required a bespoke integration deal that only large companies could afford.

Open banking is the regulatory answer to that problem. In its modern form it says: banks must expose, through standard APIs, the payment capabilities and account data that customers already have a right to — and customers can authorize licensed third parties to use those APIs on their behalf.

The two rails

Almost every open banking regime has the same two functional halves.

Account information services (AIS). A third party you authorize can read your transaction history and account balances from your bank. This powers loan underwriting, budgeting apps, income verification, and anything else that needs to know how much money actually flows through your account.

Payment initiation services (PIS). A licensed third party can start a payment directly from your bank account — no card involved. The payment is authorized in your banking app or with strong customer authentication, and it moves over real-time rails. This is the card-alternative path: cheaper for merchants, and it keeps the money in the banking system end to end.

Everything else — savings marketplaces, insurance comparison, pension dashboards — is a product built on those two primitives plus consent.

How we got here

Open banking is a regulatory export from Europe, adapted differently in each market that adopted it.

European Union / UK. The Payment Services Directive 2 (PSD2), which applied from 2019, required banks to open payment account access to licensed third parties. The UK went further operationally: the Competition and Markets Authority mandated the nine largest banks (the CMA9) to deliver a specific Open Banking standard — fixed API specifications, a directory, and a conformance regime — years before the rest of the EU converged on similar standards. The result is that the UK’s implementation is more standardized and usable than most continental ones.

United States. The US has no single federal open banking mandate. Section 1033 of the Dodd-Frank Act, which the Consumer Financial Protection Bureau finalized rules for in 2024, gives consumers the right to access and share their financial data, with phase-ins through the mid-2020s. Until those rules fully bite, the US market runs on a patchwork of screen-scraping, bank-controlled data aggregators (Plaid, Finicity, MX, and their peers), and voluntary API programs — which is exactly the fragmented mess open banking mandates were designed to end.

India. India skipped the EU playbook entirely and built something more interesting: the Account Aggregator (AA) framework, regulated by the Reserve Bank of India as a category of non-banking financial company. An AA is a licensed, consent-based data-sharing pipe — it cannot sell data, cannot keep it beyond the transaction, and cannot initiate payments. It simply moves financial information from a provider you choose to a consumer you choose, with your consent recorded at every step.

The scale is worth stating. By March 2026, official figures put the AA network at 179 live financial information providers, nearly 1,000 live financial information users, over 2.8 billion financial accounts enabled to share data, and roughly 285 million accounts actually linked by users. RBI directions issued in late 2025 formalized the framework, and the central bank has signaled interoperability among aggregators, so that your consent flows are not trapped with whichever AA a particular app chose.

Brazil. Open Finance Brazil rolled out in phases starting in 2021, covering data sharing, payments, and eventually an open insurance framework. It is currently the broadest implementation outside India in terms of data categories.

Nepal and South Asia. Nepal does not have a finished open banking data-sharing mandate. What it does have is open-API payment infrastructure: NCHL’s connectIPS and the National Payment Interface give banks, wallets, and PSPs documented APIs for real-time transfers, bill payments, and merchant payments. Most Nepali mobile banking apps and wallets (eSewa, Khalti, and others) integrate with these rails rather than each building a private network. That is open banking’s payment-initiation rail, built as public infrastructure — while the data-sharing half remains nascent.

What actually gets built on top

The infrastructure is the boring part. The products on top are where the money is.

Credit underwriting

The single highest-value use of open banking data is lending. A bank statement PDF, self-reported income, and a hope is a terrible underwriting input. Two years of transaction history from an aggregator is a much better one: actual salary credits, actual spending patterns, actual existing EMIs, actual bounce history.

This is why the AA framework in India grew fastest in lending: a lender can, with consent, pull verified transaction data in minutes instead of days of document collection, and the same data lets lenders serve people with thin formal credit files — salaried but unbanked-credit, gig workers, small traders whose income shows up as UPI credits rather than payslips.

Personal finance and account aggregation

The original consumer pitch: one app showing all your accounts. Reality check: this has been a durable but not transformative category. Aggregation is useful; it is not, by itself, a business most people will pay much for. The apps that stuck — in various markets — are the ones that turned aggregation into an action: better FX, automated savings, tax tracking, cash-flow forecasting for small businesses.

Account-to-account payments

The PIS rail lets merchants offer bank-pay at checkout as an alternative to cards. Where it has worked — the UK’s variable recurring payments for subscriptions, pay-by-bank in e-commerce — the pitch is lower cost than card interchange plus no card fraud. Where it has struggled: consumer habit. Cards have rewards, chargeback protection, and decades of muscle memory. ACH-style bank payments win where the incentive is strong enough to overcome that: large-ticket purchases, B2B, gig payouts, and markets where cards barely exist.

Income and employment verification

Rent applications, mortgage approvals, gig-work onboarding — anywhere a counterparty needs proof that money arrived in your account, open banking replaces document forgery with API calls. This is a modest category that quietly became infrastructure in the UK and US.

The hard parts, honestly

Open banking sounds like a solved problem once the regulation exists. In practice, four things break it.

Bank API quality. Mandated APIs are delivered at the minimum level banks are willing to fund. In several markets, the first years of open banking were defined by flaky endpoints, undocumented breaking changes, and authentication flows that dropped users mid-consent. Conformance testing improved this at the margins; the banks that treat APIs as a product rather than compliance overhead are still the exception.

Consent UX. Every additional consent screen loses users. A loan application that requires connecting three accounts, each with its own redirect and re-authentication, has a funnel that dies somewhere around step two. The AA model’s improvement here — one consent artifact, one aggregator, reusable across institutions — is a large part of why India’s framework scaled where older screen-scraping flows did not.

Monetization. Regulation creates the rails; it does not create willingness to pay for them. Most open banking businesses monetize on the B2B side — selling underwriting data, fraud signals, or payment initiation to institutions — rather than charging consumers for account aggregation. That is fine, but it means the consumer-facing pitch (control your data!) and the actual business (lenders buy verified data access) are different stories.

Liability and fraud. If a third party initiates a payment and the customer claims they never authorized it, who eats the loss? Open banking regimes allocate this between bank, third party, and customer under defined rules, and the details determine whether banks quietly sabotage the ecosystem by making disputed payments nearly impossible to recover. Push payments with real-time authorization are the customer’s responsibility in most regimes — the same lesson UPI learned.

What building against them is actually like

For engineers implementing against open banking APIs — as opposed to policy people designing them — a few realities are worth knowing before the first sprint.

Consent is a multi-step redirect flow, not an API key. The standard pattern: your app sends the user to the bank or aggregator’s authorization page, the user authenticates with the institution itself (never with credentials your app sees), selects accounts and permissions, and is redirected back with a short-lived authorization code that your backend exchanges for tokens. The tokens carry scopes — which accounts, which data types, what duration. Designing for this flow means handling interruption well: users abandon mid-redirect, sessions expire, and banks time out at their own pace.

Re-consent is a product problem. Consent expires — often in 90 or 180 days, depending on the regime and data type — and re-authorization is where aggregation products bleed users. The apps that handle this build reminder flows and one-tap re-consent; the ones that do not watch their data freshness quietly decay until a lender’s decisioning engine is running on stale bank data.

Sandbox ≠ production, everywhere, always. Every open banking sandbox has behaviors that differ from production institutions: different error codes, more reliable uptime, fewer institution-specific quirks. The institution-specific quirks are the real work. Two banks implementing the same standard will disagree about edge cases, pagination, transaction categorization, and how they name the same merchant. Budget for a per-institution adapter layer even when the standard says you should not need one.

Idempotency on payment initiation is not optional. PIS calls that start payments must be idempotent on your side — a retried request must not create a second payment. The rails vary in how much idempotency they enforce themselves; assuming they all do is how double-charges happen.

Rate limits and availability follow the banks, not you. Your product’s uptime ceiling is the slowest institution in your aggregation set. Design degradation paths: show cached data with clear timestamps, degrade gracefully when one bank is down, and never let one institution’s maintenance window break the whole app.

A practical map of the standards

If you are building against these systems — or evaluating them — a quick orientation:

MarketFrameworkData sharingPayment initiationNotable trait
UKOpen Banking Standard (OBIE)YesYesBest-documented APIs; strong conformance regime
EUPSD2 + Berlin GroupYesYesStandard-fragmented across national markets
USCFPB 1033 + marketEmergingEmergingStill dominated by aggregator-mediated access
IndiaAccount Aggregator (RBI)YesVia UPI rails separatelyConsent-first; payment initiation separate from AA
BrazilOpen Finance (BCB)YesYesBroadest data scope among major markets
NepalNCHL APIs / NPILimitedYesPayment rails open; data-sharing mandate nascent

Note the India row: payment initiation and data sharing live in different systems — UPI handles payments, the AA network handles data. That separation is deliberate, and arguably cleaner than the EU model where one directive created both obligations with uneven results.

Where this goes next

The phrase in circulation now is “open finance” — extending the same model beyond banks to insurance, pensions, securities, and telecom data. India’s AA framework already contemplates securities and insurance data alongside bank accounts; the UK added pensions dashboards; Brazil opened its framework to insurance. The pattern is consistent: data sharing starts with banks because banks are regulated, then spreads to every institution holding regulated financial data.

The more consequential development is quieter: payment initiation and data sharing converging with real-time rails. When a customer can authorize a bank-pay at checkout and a lender can, in the same consent flow, pull the transaction record afterward, the account — not the card — becomes the center of gravity for both payments and credit. That convergence is what the UPI-plus-Account-Aggregator stack in India demonstrates at population scale, and it is the model most other markets are circling.

For anyone building in this space, the infrastructure is no longer the bottleneck. Bank APIs exist in most markets; consent frameworks exist; regulators enforce them. What remains scarce is products that make the average person care — and institutions willing to treat the API surface as a product rather than a compliance tax. The rails are laid. The interesting work is everything that runs on them.

Frequently asked questions

Is open banking the same as open finance?
Open banking usually means bank account data and payments. Open finance extends the same consent-and-API model to insurance, pensions, securities, and other financial data. The term signals ambition more than a hard boundary.
Do banks like open banking?
Usually not initially. Mandated API access is compliance work with little direct revenue, and it hands customer relationships to third parties. Many banks have come around by building their own aggregator products and API platforms on top of the mandate.
How is my data protected under open banking?
Data only moves with explicit, time-limited consent you grant through your bank's own authentication. Third parties must be licensed and audited, and they are generally barred from storing data longer than needed or from using it for purposes beyond what you consented to. Regulators enforce the rules with fines.
Will Nepal get an open banking framework?
Nepal already has open-API payment infrastructure through NCHL's connectIPS and the National Payment Interface, and its wallets and mobile banking apps build on those APIs. A fuller data-sharing framework on the Account Aggregator model remains a policy work in progress rather than a finished system.