For banks, fintechs and business-banking platforms
Make business banking more than banking.
Your platform knows where your customers' money moved. Finanly is the layer that explains what it meant: reconciled accounting, reporting and governed intelligence on top of the account activity you already hold.
Nine problems innovation teams actually have
Written for the roadmap, not the brochure.
Each card states the problem as banks describe it, then what Finanly contributes — and where it stops. We would rather be precise than impressive.
The relationship ends at the transaction
A bank can show where money moved. Its business customers still leave for separate accounting, analytics and commerce tools to understand what it meant.
Finanly turns account activity into categorized, reconciled, explained financial records — the part of the relationship that happens after the payment.
Engagement that payments alone can't create
Balances and transfers are checked; they are not worked in. Financial management is where owners spend their attention every week.
A close queue, reconciliation, statements and a question box give customers reasons to return. A strategic opportunity, not a retention claim we have measured for you.
Commoditized business banking
Cards, payments, balances and sub-accounts are table stakes. Differentiation has moved to what the platform understands about the business.
Context-aware financial intelligence on top of banking data: margin by product line, payout decomposition, cash runway — answered from governed measures, not guessed.
Fragmented financial data
Owners run money across banks, processors, stores, freight, currencies and an accounting system. No single party sees the whole picture.
Eight live connectors normalize those sources into one canonical model; the same engine reconciles a Shopify payout, a freight bill and an FX revaluation.
The cost of building it yourself
A reconciliation and intelligence stack is many layers: connectors, a data model, accounting workflows, a semantic layer, controls, a model layer, a UI.
Finanly's modular stack already exists and runs four companies' books. Integration still takes work; starting from zero takes years.
Richer, permissioned visibility
Institutions see cash in and out. They rarely see receivables, payables, inventory or margin — the context behind a business's health.
Permissioned, reconciled financial context could inform more useful services. Finanly does not perform credit underwriting or make credit decisions.
AI that must not invent numbers
A chatbot that hallucinates a balance is a regulatory and reputational problem, not a feature.
The model proposes a plan; the platform validates it; the engine computes; a human approves postings. Every answer carries its plan and its rows.
Enterprise control
Tenant isolation, permissions, traceability, deployment options, vendor risk — the questions that decide whether anything ships.
Row-level security per company, encrypted secrets, append-only audit, idempotent side effects, self-hosted deployment. Documented controls; SOC 2 not yet certified.
Commercial upside
Subscription features, higher-value services and differentiated experiences are the business case — but only if the product is real.
Plan-gated feature entitlements and billing are built in. We won't project revenue or retention for you; we will show you the product doing the work.
A bank statement is a list of facts. A business needs an account of itself. The institution that provides the second will own more of the relationship than the one that only provides the first.
What your customers could see
Seven product opportunities, each already running somewhere.
Everything below exists in the product today for the companies that run on Finanly. Making it part of your platform is integration work, not invention.
Embedded accounting and financial management
Categorization with review, splits and undo; a month-end close with an exception queue; invoicing and recurring invoices; exports to a ledger of record.
Transaction enrichment and reconciliation
Counterparty enrichment, three-tier categorization, transfer matching, payout and fee decomposition, freight billing, dispute tracking — to the cent.
Reporting and analytics
Balance sheet, P&L and trial balance in parity with the ledger; Report Studio and dashboards on a governed semantic layer; exports in every common format.
Natural-language questions grounded in accounting data
Plain-English questions answered from supported measures, with the plan and available source detail. Scope depends on your connected data, permissions and reviewed definitions.
Contextual visibility across systems
Commerce, logistics, analytics and ledger data joined to banking data: margin, inventory, receivables, payables, cash — one account of the business.
Multi-entity experiences
Several companies under one login, isolated by row-level security, with roles per company and plan-gated features — plus a Personal mode on the same engine.
Institution-controlled deployment
One Docker Compose stack on infrastructure you control, with the model layer on the same host. Private network for everything but the API and UI.
AI reliability
Four hands, hard limits.
The separation your risk team will ask about: language understanding, validation, calculation and approval are different components with different permissions.
The model proposes
Reads the request and the governed vocabulary; returns a semantic plan or a categorization suggestion with a confidence.
Never reads raw ledger rows. Never calls a connector.
The platform validates
Checks every measure, dimension and period against the registry; binds the tenant from the session; fails closed on anything unknown.
Never trusts a field the model invented. Never widens a tenant boundary.
The engine calculates
Executes the validated plan on the canonical store and returns rows, totals and the plan that produced them.
Never computes outside governed measures. Never returns a figure without its rows.
A human approves
Accepts, edits, splits or rejects; approves export batches to the ledger. Every decision is reversible and audited.
Nothing irreversible happens without this step.
Honest availability
Available today. Needs integration work. Planned.
The matrix we would want to read before a first call. Statuses are maintained against the platform's feature registry and the running stack.
Financial management experiences
- Transaction categorization with review, split, undoThree-tier AI Match; human-in-the-loopAvailable today
- Month-end close with exception queueAudited, idempotent exports to a ledger of recordAvailable today
- Payout reconciliation and fee decompositionShopify today; same engine pattern for other processorsAvailable today
- Invoicing, recurring invoices, branded PDFLedger of record: ERPNextAvailable today
- Multi-currency accounts and FX revaluationApproval-gated exportAvailable today
Intelligence & reporting
- Natural-language questions on governed measuresSupport depends on reviewed measures, available data and workspace accessAvailable today
- Statements (BS, P&L, TB) with hourly parityPer company, per periodAvailable today
- Report Studio and dashboardsCSV / XLSX / PNG / PDF exportsAvailable today
- Scenario narrative over run-rateNarrative, not a certified forecasting engineAvailable · configuration
- Board-pack generatorNot builtPlanned
Platform & integration
- Multi-company tenancy, RBAC, feature entitlementsPostgres RLS; plan-gated feature keysAvailable today
- Self-hosted deployment (Docker Compose, one host or VPS)Private model and gateway on the same hostAvailable today
- Bank feeds via aggregator or direct APIPlaid, Mercury, Wise, PayPal liveAvailable today
- Institution's own core-banking feed as a connectorConnector contract exists; a new service per sourceNeeds integration work
- Embedding Finanly screens inside a bank's appSame-origin web app today; embedded SDK not builtNeeds integration work
- QuickBooks Online / Xero as ledger of recordCapability matrix defined; not implementedPlanned
- SOC 2 reportControls in place and documented; not yet certifiedPlanned
Architecture for integration
One public API. Everything else private.
The browser — or your platform — only ever talks to core-api. Connectors, the model gateway, workers and the compliance engine live on a private network and are called with short-lived service tokens.
APIs
A versioned REST API (OpenAPI) behind the edge. Every side-effect endpoint takes an Idempotency-Key; every response carries a request ID.
core-api · /v1/*Data boundaries
Each connector holds only its own credentials, encrypted with a tenant-bound key. The model gateway has no host port and sees one tenant per call.
AES-GCM · internal JWTTenancy
Each company is a tenant. Row-level security in Postgres enforces isolation underneath the application; memberships carry roles; features are plan-gated.
RLS · memberships · plan_featuresControls
Append-only audit events (database trigger), approvals and export batches, rate limits on public routes, undo for categorizations and postings.
audit_events · export_batchesConnectors
Eight live services with one internal contract: fetch and normalize, never persist. A new source — including an institution's own feed — is a new service, not a rewrite.
connector-* · capability matrixDeployment
One Docker Compose stack on a host or VPS you control; Postgres and Redis on local volumes; the model served on the same host. Rollback by image tag.
compose.core.ymlHow a partnership starts
Three steps, no deck required.
- A first conversation
You describe the customer segment and the gap. We walk through the product on demo data and share the architecture and availability matrix — the same ones on this page, in more depth.
- A technical review
Your engineering and risk teams get the API surface, the deployment model, the data-boundary and tenancy design, and the model-layer contract. Questions answered in writing.
- A scoped integration
A defined slice — one data source, one workflow, one customer cohort — on infrastructure you control, with explicit acceptance criteria. Timelines are agreed then, not promised here.
Discuss a technology partnership
Tell us what your customers are missing.
A direct line to the people who build Finanly. We reply personally, usually within two working days, with a technical overview and a proposed first call. No customer data, no credentials — just the problem.
- Platform integrations and embedded financial management
- Technology partnerships and deployment reviews
- Questions about the architecture, the model layer or the controls
FAQ
Straight answers for platform teams
Is Finanly a white-label product?
Not today. It is a self-hosted web application with an API-first core. Embedding screens inside another app is integration work we would scope together; an embeddable SDK is not built.
Can it run inside our environment?
Yes — that is the default. One Docker Compose stack on infrastructure you control, with the model layer on the same host. We do not operate a multi-customer cloud.
Which ledger does it post to?
ERPNext today, through a dedicated connector. The ledger connector contract defines a capability matrix; QuickBooks Online and Xero are planned, not shipped.
How does the AI avoid inventing figures?
It only emits a semantic plan. core-api validates every measure, dimension and period against the governed registry; the engine computes the numbers; unknown fields fail closed.
Does Finanly make credit decisions?
No. It produces reconciled, permissioned financial context. Any use in lending or underwriting would be your decision, your models and your regulatory responsibility.
What about SOC 2 and certifications?
Finanly is not SOC 2 certified. The controls — forced row-level security, encrypted secrets, append-only audit, internal-only AI, idempotent side effects — are in place and documented, and we will walk your team through them.
What proof of operation exists?
Four operating companies have closed their books on the platform since February 2026, across two currencies, with commerce and freight reconciliation and statements kept in parity with the ledger. We show the real system on demo data.
How do we start?
Use the form above. You will get a short technical overview and a proposed first call; a scoped integration follows only if it makes sense for both sides.