Use case

AI tool transaction enrichment

Let ChatGPT, custom GPTs, coding agents, and MCP-capable assistants call one safe enrichment endpoint instead of guessing merchant metadata from raw descriptors.

Answer-first summary

TxnKit fits AI workflows that already have a safe transaction descriptor and need a structured merchant enrichment result. The AI tool should import the OpenAPI contract, call `POST /v1/enrich`, and treat confidence, signals, and warnings as constraints instead of filling gaps with invented merchant facts.

Raw descriptor example

POST /v1/enrich

{
  "raw_description": "SQ *JOES COFFEE 0421 TORONTO",
  "amount": 6.75,
  "currency": "CAD",
  "country": "CA",
  "mcc": "5814"
}

{
  "display_name": "Joe's Coffee",
  "category": "Food and Drink",
  "processor_hint": "square",
  "confidence": 0.72,
  "signals": ["removed_square_prefix", "detected_city_token"],
  "warnings": ["local_merchant_identity_not_verified"]
}

ChatGPT and custom GPT setup

Use the public OpenAPI file as the action schema, configure API key auth when public keys are available, and expose only the enrichment operation. The tool description should say that the action accepts one display-safe descriptor and rejects sensitive financial or personal data.

When to use

Use this when an AI assistant is helping build or inspect a transaction-feed feature and needs a deterministic merchant display result with confidence-aware fallbacks.

When not to use

Do not use this flow for uploaded statements, batches of transactions, card numbers, account numbers, bank credentials, contact details, customer names, customer PII, or autonomous financial decisions.