Skip to main content
Every release of the ParseTx API is documented here in reverse chronological order. You’ll find new features, bug fixes, and architectural changes for each version. Breaking changes are always called out explicitly so you can plan upgrades with confidence.

​
v0.1.0 — Public Beta Launch

Released June 2026 — This is the initial public release of the ParseTx API. If you’re onboarding now, you’re building on the full feature set described below.

​
✨ New Features

  • Synchronous transaction enrichment — POST /v1/enrich accepts up to 10 transactions per request and returns structured merchant data with sub-50ms response times for cache hits. Send a messy bank descriptor; get back a clean merchant name, domain, category, MCC code, subscription flag, and confidence score.
  • Asynchronous batch enrichment — POST /v1/enrich/async accepts up to 500 transactions per request via a background job queue. The endpoint immediately returns a job_id with a 202 Accepted response. Poll GET /v1/enrich/jobs/:id to retrieve results when processing completes.
  • 15-category merchant taxonomy — Every enriched transaction is classified into one of 15 standardized categories: Shopping, Groceries, Food & Drink, Gas & Fuel, Transportation, Entertainment, Digital Services, Software, Utilities, Travel, Transfer, Payroll & Tax, Healthcare, Fees & Charges, and Unknown. Categories are consistent across all requests.
  • PII sanitization pipeline — All transaction strings are sanitized before leaving your infrastructure boundary. The pipeline strips numeric sequences longer than four digits (card numbers, account numbers, routing numbers), detects and redacts personal names embedded in payment processor strings (e.g. SQ *JANE DOE PORTLAND OR), and truncates inputs to 64 characters. Raw transaction strings are never logged — only anonymized hashes.
  • Cost-aware AI gateway — Cache misses are routed through a paid-tier AI inference pipeline governed by Google’s Data Processing Addendum, which prohibits the use of request data for model training. Responses are validated against a strict schema before being written to cache, ensuring structural consistency on every result.
  • Official Node.js SDK — Install with npm install @parsetx/sdk. Supports both synchronous and asynchronous enrichment with full TypeScript types.
  • Official Python SDK — Install with pip install parsetx. Compatible with Python 3.9+.
  • Pay-as-you-go pricing via Stripe Metered Billing — A flat $0.005 per transaction, billed monthly based on actual consumption. No monthly minimums, no tier upgrades, no contracts. A credit card is required at sign-up to generate an API key.

​
🏗️ Infrastructure

  • Global edge cache pre-warmed with 166 top consumer merchants, giving immediate sub-50ms responses for the most common transaction descriptors on day one.
  • Persistent cache backup layer ensures results survive edge cache eviction with a 30-day retention window.
  • Subscription state changes are validated in real time, with HMAC SHA-256 signature verification and a five-minute replay-attack window.
  • Per-request 10-second timeout on all AI inference calls — if the upstream model is slow, affected items return with status: "retry" rather than blocking the entire batch.

​
🐛 Bug Fixes

  • Fixed an issue where AI inference calls returned 404 errors due to an incorrect upstream API path.
  • Fixed a Square merchant regex that was incorrectly stripping single-word merchant names (e.g. Grubhub) to an empty string.
  • Fixed a webhook handler crash caused by a missing column in the key management table.
You’re on the latest version. Watch this page for upcoming releases — new enrichment fields, SDK updates, and performance improvements are on the roadmap.