Skip to main content

Product

From raw description to structured context.

TXNORA is an API for automated data processing with a focus on transactions: a raw transaction becomes a recognizable merchant, a category, contextual labels and a confidence value. The same platform also categorizes, classifies and extracts information from other data.

Request trial accessDeveloper overview

How it works

Five steps from raw data to context.

One transaction as the example: this is how a raw description becomes a structured result.

  1. Take in raw data

    It starts with a transaction as it exists in your system: counterparty, payment reference, amount, currency and booking date. Your product sends this raw data to the API in a request.

    Raw data

    Counterparty
    PAYPAL *ADOBE
    Payment reference
    ADOBE SOFTWARE
    Amount
    −59.49 EUR
    Booking date
    Oct 5, 2026

  2. Recognize the merchant

    Transaction descriptions are often truncated and include payment providers, store numbers or references. TXNORA derives a recognizable merchant name from them.

    Counterparty
    PAYPAL *ADOBE
    merchant.name
    Adobe

  3. Assign a category

    The merchant alone does not say what was paid for. So TXNORA assigns the transaction a category from one consistent structure, with a stable ID for your logic and a readable name for display.

    merchant.name
    Adobe
    category
    Softwaresoftware

  4. Add contextual labels

    A category rarely answers every question. Labels add further context, for example that this is a digital service. A transaction can carry several labels.

    category
    Softwaresoftware
    labels[]
    • SOFTWARE
    • DIGITAL_SERVICE

  5. Return confidence

    TXNORA returns a confidence value between 0 and 1 with every result. Your product decides, by its own rules, what to accept directly and what to send to review.

    merchant.name
    Adobe
    confidence
    98.4%

Capabilities in depth

Four building blocks, one result.

Each building block answers its own question about a transaction. Together they form the structured context your product works with.

  • Who was paid?

    Merchant recognition

    Payment providers, marketplace codes and reference numbers often hide the actual merchant in the transaction description. TXNORA turns them into a clear, recognizable name. Your product can then show a merchant instead of a string of characters.

    Examplesmerchant.name

    • PAYPAL *ADOBE
      →Adobe
    • REWE SAGT DANKE 1834
      →REWE
    • AMZN MKTP DE*X4H88291
      →Amazon

  • What for?

    Transaction classification

    Each transaction is assigned a category from one consistent structure. The category has a stable ID for logic and filters and a name for display. This lets your product group and analyze transactions consistently.

    Examplescategory

    • Adobe
      →Softwaresoftware
    • REWE
      →Groceriesgroceries
    • Amazon
      →Retailretail

  • What else?

    Contextual labels

    Labels describe what a single category does not cover, such as a digital service or online retail. A transaction can carry several labels. Your product can use them for filters, hints or its own analyses.

    Exampleslabels[]

    • Adobe
      →
      • SOFTWARE
      • DIGITAL_SERVICE
    • REWE
      →
      • GROCERIES
      • RETAIL
    • Amazon
      →
      • ECOMMERCE
      • RETAIL

  • How certain?

    Confidence scores

    TXNORA returns a confidence value with every prediction. Your product sets its own threshold for when a result is accepted directly and when it goes to review. The value refers to the individual result, not to a measured overall accuracy.

    Examplesconfidence

    • Adobe
      →98.4%
    • REWE
      →99.1%
    • Amazon
      →96.2%

Examples from three typical transactions. Categories and labels appear exactly as they come back in the response.

Output fields

What a response contains.

The response is deliberately lean: a few clearly named fields. Choose an example to compare the values.

Choose an example

CounterpartyPAYPAL *ADOBEREWE SAGT DANKE 1834AMZN MKTP DE*X4H88291

merchant.name
Type: string
Recognized merchant in readable form.
Example value"Adobe""REWE""Amazon"
category.id
Type: string
Stable category identifier for logic, filters and analyses.
Example value"software""groceries""retail"
category.name
Type: string
Display name of the category.
Example value"Software""Groceries""Retail"
labels[]
Type: string[]
Additional context; a transaction can carry several labels.
Example value["SOFTWARE", "DIGITAL_SERVICE"]["GROCERIES", "RETAIL"]["ECOMMERCE", "RETAIL"]
confidence
Type: number (0–1)
Confidence of the result. Your product decides what gets reviewed.
Example value0.9840.9910.962

Simplified example of a response.

See request and response

Scope

What TXNORA is not.

To set clear expectations: TXNORA is a data building block for your software, not a finished application or an advisory service.

  • Not a banking app

    TXNORA offers no interface for end customers, holds no accounts and initiates no payments. The API provides context that your product brings into its own interface.

  • Not a tax or accounting service

    TXNORA does not give tax advice and does not keep books. Structured data can support review and bookkeeping workflows in your product, but it does not replace a professional review.

  • No tax assessment of individual transactions

    TXNORA does not derive tax deductibility, input VAT, business use or recurring payments from a merchant name, category or labels. A label such as SOFTWARE describes context, not tax treatment.

Instead: a clearly defined data building block. What your product does with it is your decision.

The platform

Seven functions for unstructured data.

Transactions are the focus. The same platform also processes text, documents and records from other domains – through simple REST endpoints.

  • Categorization

    Automatically assigns text, documents or records to categories.

    ExampleSupport ticket

    I received the invoice for March twice.

    →
    Category
    Billing

  • Labeling

    Automatically tags data with the labels you define.

    ExampleTransaction

    PAYPAL *ADOBE

    →
    • SOFTWARE
    • DIGITAL_SERVICE

  • Data Annotation

    Annotates large volumes of data in a structured way, such as statements, attributes or sentiment in text.

    ExampleProduct review

    Fast delivery, but the battery doesn’t last long.

    →
    Delivery
    positive
    Battery
    negative

  • Data Classification

    Classifies data automatically according to your own specifications.

    ExampleEmail

    Please cancel my contract at the end of the month.

    →
    Class
    Cancellation

  • Information Extraction

    Extracts relevant information from unstructured data and returns it as fields.

    ExamplePayment reference

    INVOICE INV-2026-0412 DATED 10/01

    →
    Invoice number
    INV-2026-0412

  • Data Intelligence

    Analyzes, structures and enriches existing data, for example inconsistent master data.

    ExampleMerchant data

    AMZN MKTP DE · Amazon.de · AMZN Mktp

    →
    Merchant
    Amazon

  • Custom Endpoints

    Custom processing based on your prompts, rules and data structures – delivered as your own endpoint.

    More on Custom Endpoints

    ExampleDelivery note

    Delivery for order PO-2026-118, item 2 not delivered.

    →
    order_reference
    PO-2026-118
    needs_review
    true

Infrastructure & data protection

GDPR-compliant, hosted locally or in the EU.

TXNORA processes your data on LLM endpoints hosted locally or in the EU. Sensitive company data is not passed to conventional public AI services, and you do not need your own LLM infrastructure.

Data flow
Your system
TXNORAREST API
LLM endpointshosted locally or in the EU
Public AI servicesno sensitive data passed on

  • Hosted locally or in the EU

    The LLM endpoints that TXNORA uses to process your data run locally or in the EU.

  • GDPR-compliant processing

    Processing is GDPR-compliant. If you have questions about the data flow in your case, include them in your request.

  • No public AI services

    Sensitive company data is not passed to conventional public AI services.

  • No LLM infrastructure of your own

    You do not run models or servers for them. TXNORA provides the processing as a service.

  • Simple REST API

    Your system sends data in a request and receives structured results as JSON. That keeps integration fast.

  • Scalable processing

    From single requests to large volumes of data: processing grows with your needs.

Custom endpoints

Your endpoint, your rules.

When standard functions are not enough: a custom endpoint processes data based on your own prompts, rules and data structures – for special data models, business processes or large data volumes.

  • Your own models and prompts

    Tailored to your domain and the terms your team works with.

  • Your own rules

    Requirements and review rules from your processes feed directly into processing.

  • Your own data structures

    The response follows your data model, with exactly the fields your system expects.

  • Large data volumes

    For single requests as well as large datasets that need automated processing.

We work out what your endpoint needs together after your request.

Example: checking a delivery note

Input

Delivery for order PO-2026-118, item 2 not delivered.

Your rules

  • Detect the order number
  • Flag missing items
  • Review on mismatch

Response in your schema

{
  "order_reference": "PO-2026-118",
  "missing_items": [2],
  "needs_review": true
}

Pricing

Usage-based and transparent.

You pay for the processing you actually use – on very competitive terms.

You receive the billing unit and pricing together with your trial access.

Request pricing

  • Usage-based

    Billing follows usage. You do not have to fund an LLM infrastructure of your own.

  • Transparent

    Clear terms that you know before you start.

  • Competitive

    An alternative to international LLM APIs, with processing locally or in the EU.

Next step

Is TXNORA a fit for your product?

Find out with free trial access. Within 24 hours of your request, you receive access to the API, the developer documentation and our pricing.

Request trial accessSee use cases