The answers your team needs before going live.

Most teams send their first categorised transaction in under 2 hours using the Python or TypeScript SDK. Full production integration — including TrustFlow pipeline setup, custom taxonomy configuration, and reconciliation rules — typically takes 2–6 weeks depending on the complexity of your data sources. Our sandbox environment mirrors production behaviour so you can test thoroughly before going live.

No. The Starter plan is permanently free for up to 5,000 transactions per month and requires no credit card. You get full API access, a sandbox with production-accurate data models, and the Python and TypeScript SDKs. Many teams complete their evaluation and integration entirely on the free tier before upgrading.

Yes. Every account — including Starter — includes access to the sandbox environment. The sandbox uses the same API contracts, the same model versions, and the same data schemas as production. It's loaded with realistic synthetic transaction data across common UK financial sectors. There are no surprises when you move to production.

An API key and five minutes. The quick-start guide takes you from key generation to your first categorised transaction in under 10 steps. SDKs are available for Python, TypeScript, Go, Rust, and Ruby. The full OpenAPI 3.1 spec is available without authentication.

95–99% accuracy out of the box, depending on transaction type and sector. UK payment and bank feed data performs at the higher end. The model has been trained on anonymised UK financial transaction data across payment institutions, lending platforms, wealth managers, and e-money institutions. Custom taxonomy support is available on Growth and Enterprise plans.

Yes. Growth and Enterprise plans support custom taxonomy configuration. You define your own category tree — as deep as you need — and the categorisation model maps to your taxonomy rather than the default Trustinera AI taxonomy. Custom models trained on your specific transaction corpus are available on Enterprise plans.

Every categorisation decision includes a confidence score. You configure the confidence threshold below which items surface for human review rather than being automatically categorised. This means your team spends time only on the genuinely uncertain items — typically 2–5% of volume — rather than manually reviewing the 95–98% the model gets right.

Each API response includes: the assigned category, a confidence score (0–1), the model version that made the decision, the top contributing factors, and a lineage chain from raw input to category output. This is the data your auditors, compliance teams, and regulators will ask for — it's there by default on every response.

Yes, on Enterprise plans via TrustFlow. You can register your own model, define its input schema and output format, evaluate it against benchmark datasets, and deploy it with promotion gates and full versioning. Trustinera AI handles inference infrastructure, monitoring, and the audit trail — you control the model.

By default, all data is processed and stored in UK data centres. EU residency is available on request at no additional cost on Growth and Enterprise plans. On Enterprise plans, on-premise and hybrid deployment means your data never leaves your own infrastructure. We do not share data between customer environments under any circumstances.

Yes. Trustinera AI holds SOC 2 Type II certification. Our security controls, data handling procedures, and incident response processes are independently audited annually. The SOC 2 report is available to prospects under NDA and to customers on request.

Yes, on Enterprise plans. Trustinera AI can be deployed on your own Kubernetes infrastructure using our published Helm charts. In on-premise mode, no data leaves your environment. The control plane remains cloud-hosted for licensing and model update delivery; all data processing and storage is within your perimeter.

Trustinera AI is designed for GDPR compliance. We process personal financial data as a data processor under your instruction; our Data Processing Agreement (DPA) is available for all paid plans. You can configure data retention periods, fields to be anonymised, and regional storage preferences. Transaction data is processed with the minimum identifiers necessary for categorisation — we don't require PII for most operations.

Trustinera AI is FCA-aware — meaning it's built with FCA supervisory expectations in mind. It does not hold FCA authorisation itself (it is infrastructure, not a regulated activity). However, the platform provides the documented decision trails, AML screening records, and audit pack generation that FCA-supervised firms need to meet their obligations. Our compliance page documents the specific FCA frameworks our customers use us to support.

Growth (£499/month) covers up to 100,000 transactions per month, 5 users, standard integrations (Xero, Sage, QuickBooks), and the core modules (Categorise, Reconcile, OpsIQ). Enterprise (from £2,500/month) covers unlimited volume, all modules including Sentrise and TrustFlow, on-premise deployment, SLA-backed support, custom models, and dedicated customer success.

Yes. Annual billing receives a 20% discount versus monthly — Growth plan drops from £499/month to £399/month (billed annually at £4,788). Contact us for annual Enterprise pricing.

Absolutely. You can export all your transaction records, categorisation decisions, reconciliation records, and audit logs at any time in standard formats (CSV, JSON, Parquet). There are no lock-in mechanisms. On Enterprise plans, data portability support is included in the offboarding process.

Growth plan includes email support with responses within 8 business hours, access to the full developer documentation, API changelog and deprecation notices (90-day minimum advance notice), and access to the customer community forum. Enterprise plans include a named customer success manager and priority support SLA.

Yes. Accounting platforms, banks, and embedded finance providers can deploy Trustinera AI under a white-label agreement. This includes custom branding, API key management per sub-customer, revenue share models, and joint go-to-market support. Contact us at partner@trustinera.ai.

Native bidirectional integrations: Xero, QuickBooks Online, Sage 200, Sage Accounting, and FreeAgent. CSV/SFTP import and export is available for other packages.

Currently supported via open banking and direct API connections: Barclays Business, HSBC Business, Lloyds Commercial, NatWest Business, Starling Bank, Monzo Business, and Revolut Business. Open Banking (FCA-registered) connections cover most UK current account providers. SFTP/CSV feed support is available for institutions not yet on the direct integration list.

Yes. Enterprise deployments use published Helm charts and support any CNCF-conformant Kubernetes distribution. The platform has been tested on EKS, GKE, AKS, and on-premise k8s (Rancher, OpenShift). Prometheus metrics are natively exported; dashboards are available as Grafana templates.

Via the Corrections API. When a user or your team corrects a categorisation, you POST the transaction ID, the original category, and the correct category. Corrections are used in the continuous training pipeline and improve accuracy across your account. On Enterprise plans with custom models, corrections feed exclusively into your model — they don't influence other customers' accuracy.