# ParseRail

> The AI back-end for your product.

ParseRail is the developer platform from Compound. It parses documents, extracts fields, redacts PII, analyzes contracts, fights chargebacks and enriches companies — each capability solved once and running in production all day, with Compound Labs as customer zero. 39 of those capabilities are open to developers as AI endpoints behind one API key and one pay-per-call credit wallet. No subscription, no per-seat pricing; failed calls are never charged.

## How it works
- Base URL: https://api.thecompound.tech
- Auth: `Authorization: Bearer ksk_live_…` (mint a key in the dashboard)
- Pricing: 1 credit = $0.01. Each endpoint burns credits at its own rate; you are only charged on a successful call. Buy credits in the dashboard.
- Response envelope: every successful call returns its result plus `usage: { credits, balanceRemaining }`.
- Errors: a stable `{ error: { code, message } }` envelope.

## Endpoints
- POST /v1/parse — Document parse. Any invoice, receipt, EOB, ERA, or COI — PDF or image — into structured, validated JSON. (10 credits · $0.10/call)
- POST /v1/extract — Field extraction. Pull a field set you define out of any block of text. You name the fields; you get typed values with confidence. (4 credits · $0.04/call)
- POST /v1/classify — Classification. Route or tag text against your own taxonomy — a label, a confidence, and a one-line rationale. (2 credits · $0.02/call)
- POST /v1/summarize — Summarize. A meeting transcript, thread, or report → a tight summary, key points, and extracted action items. (3 credits · $0.03/call)
- POST /v1/redact — PII redaction. Detect and strip names, emails, phones, SSNs, cards, and PHI from text before you store or log it. (3 credits · $0.03/call)
- POST /v1/sentiment — Sentiment. Turn a review or support message into a sentiment score, per-aspect breakdown, and the themes driving it. (2 credits · $0.02/call)
- POST /v1/contract — Contract analysis. A contract → parties, effective date, term, renewal, governing law, obligations, and flagged risk clauses. (12 credits · $0.12/call)
- POST /v1/chargeback — Chargeback representment. Dispute details → a representment narrative, an evidence checklist, the right reason code, and a win-likelihood. (15 credits · $0.15/call)
- POST /v1/enrich — Company enrichment. A domain or work email → a structured company profile: name, description, industry, HQ, size, and links. (5 credits · $0.05/call)
- POST /v1/invoice — Invoice extraction. An invoice — PDF, photo, or text — into vendor, dates, PO refs, tax, totals, and clean line items. (8 credits · $0.08/call)
- POST /v1/receipt — Receipt extraction. A receipt into merchant, items, totals, payment method, and an expense category — built for expense flows. (6 credits · $0.06/call)
- POST /v1/statement — Statement parsing. A bank or card statement into the account, the period, balances, and every transaction as a normalized row. (12 credits · $0.12/call)
- POST /v1/resume — Resume parsing. A resume or CV into a structured candidate profile: contact, skills, experience, education, and links. (8 credits · $0.08/call)
- POST /v1/tables — Table extraction. Every table in a document — even scanned — as clean headers and rows, ready for your spreadsheet or DB. (8 credits · $0.08/call)
- POST /v1/split — Document splitting. A multi-document scan bundle classified and split: what each document is, where it starts and ends, and a summary. (10 credits · $0.10/call)
- POST /v1/compare — Document comparison. Two versions of a contract or document → every material change, what it means, and the risk it carries. (15 credits · $0.15/call)
- POST /v1/structure — Structure to your schema. Any messy input — text, HTML, an email — plus YOUR JSON schema → output shaped to it, validated against your required fields and property types, with an automatic corrective retry and a `valid` flag. (6 credits · $0.06/call)
- POST /v1/normalize — Record normalization. A batch of messy records → clean canonical rows, with a change log of every fix (casing, formats, dedup-ready values). (5 credits · $0.05/call)
- POST /v1/match — Entity matching. Two record sets → which rows are the same real-world thing, with confidence and reasoning. Fuzzy names, typos, aliases handled. (8 credits · $0.08/call)
- POST /v1/categorize — Batch categorization. Up to a hundred items against your taxonomy in one call — products, transactions, tickets — each with a confidence. (3 credits · $0.03/call)
- POST /v1/dunning — Collections sequence. An overdue invoice → a ready-to-send collection sequence, escalating at the right pace and tone for how late it is. (8 credits · $0.08/call)
- POST /v1/po-match — 3-way match. Invoice vs purchase order vs receipt → matched, partial, or mismatched, with every discrepancy flagged and sized. (10 credits · $0.10/call)
- POST /v1/quote — Quote builder. A job description plus your rates → an itemized, professional quote with assumptions and scope notes spelled out. (8 credits · $0.08/call)
- POST /v1/fraud-flag — Fraud triage. An order or transaction in context → a risk score, the signals driving it, and the checks worth running before you ship. (6 credits · $0.06/call)
- POST /v1/reply — Reply drafting. A customer thread plus your context → a ready-to-send reply in the right tone, with an internal note for the agent. (5 credits · $0.05/call)
- POST /v1/triage — Ticket triage. A support ticket → priority, category, the team it belongs to, sentiment, SLA risk, and a suggested first response. (3 credits · $0.03/call)
- POST /v1/minutes — Meeting minutes. A meeting transcript → clean minutes: summary, decisions made, action items with owners, and open questions. (6 credits · $0.06/call)
- POST /v1/outreach — Outreach sequence. An enriched lead plus what you sell → a personalized outreach sequence that references what actually makes them a fit. (6 credits · $0.06/call)
- POST /v1/review-reply — Review response. A customer review → a brand-safe public response, the issues to log, and whether it needs human escalation. (3 credits · $0.03/call)
- POST /v1/moderate — Content moderation. Text or an image against YOUR policy → allow, review, or block, with the categories and excerpts that drove the call. (2 credits · $0.02/call)
- POST /v1/rewrite — Brand-voice rewrite. Any copy → your brand voice. Describe the voice or paste a sample; get the rewrite and what changed. (3 credits · $0.03/call)
- POST /v1/product-copy — Product copy. Specs and an audience → listing-ready titles, bullets, a description, and SEO keywords — per channel. (5 credits · $0.05/call)
- POST /v1/describe — Image description. An image → accessible alt text, a caption, tags, and any text found inside it. Accessibility and catalogs in one call. (3 credits · $0.03/call)
- POST /v1/research — Research brief. A company or topic → a multi-source, citation-backed brief: what it is, what changed lately, and what matters. Real web work. (50 credits · $0.50/call)
- POST /v1/screen — Lead screening. A lead plus your ICP criteria → a web-grounded qualification verdict with the evidence for and against. (10 credits · $0.10/call)
- POST /v1/transcribe — Transcription. An audio recording → accurate text with speakers and paragraph timestamps. Meetings, calls, voice notes. (10 credits · $0.10/call) [PAUSED since 2026-08-19 — returns 503, never charged]
- POST /v1/memory — Agent memory. Store, search, and forget memories for your agents — semantic recall on your own namespace, no vector DB to run. (2 credits · $0.02/call)
- POST /v1/image — Image generation. A prompt → a production-ready image. Marketing visuals, product scenes, and consistent brand imagery. (10 credits · $0.10/call)
- POST /v1/speak — Text to speech. Text → natural speech audio, ready to embed. Voice notes, IVR lines, narration. (5 credits · $0.05/call)

## Credit packs (one-time, pay-as-you-go)
- Starter — $20: 2,000 credits
- Growth — $100 (+5% bonus): 10,500 credits
- Scale — $500 (+10% bonus): 55,000 credits

## SDKs & tools
- TypeScript / JavaScript SDK: `npm i @compound/api` — https://www.npmjs.com/package/@compound/api
- Python SDK: `pip install compound` — https://pypi.org/project/compound/
- MCP server (Claude, Cursor, any MCP client): `npx @compound/api-mcp` — https://www.npmjs.com/package/@compound/api-mcp

## Links
- Docs (quickstart, per-endpoint guides, async & webhooks): https://parserail.thecompound.tech/docs
- Home + interactive reference (try it live): https://parserail.thecompound.tech
- Invoice OCR API pricing compared — per-document rates and minimums across ParseRail and ten other vendors, and the free tiers those vendors run, each read at the vendor's own pricing page and dated: https://parserail.thecompound.tech/guides/invoice-ocr-api-pricing
- How much does an invoice OCR API cost per month? (2026). The monthly bill for 500, 2,000 and 10,000 invoices a month across AWS Textract, LlamaParse, Mindee, ParseRail, Google Document AI, DocuClipper, Veryfi and Azure Document Intelligence, with every minimum and plan floor folded into the figure. Includes the volume at which Veryfi's $500/mo minimum stops binding (3,125 invoices) and which vendors publish no price at all above their free band: https://parserail.thecompound.tech/guides/invoice-ocr-api-monthly-cost
- The best document extraction APIs in 2026 — which API fits which job (shipping fast, bank statements, high volume, AWS-native, no commitment), with the vendor terms each verdict rests on and the date each was read: https://parserail.thecompound.tech/guides/best-document-extraction-api
- Receipt OCR APIs for developers, compared (2026) — per-receipt rates, entry prices and minimums across Taggun, Tabscanner, Veryfi, Nanonets, AWS Textract, Google Document AI and ParseRail (which runs no free tier), each read at the vendor's own pricing page and dated: https://parserail.thecompound.tech/guides/receipt-ocr-api
- Free invoice and receipt OCR APIs: 16 free tiers compared (2026) — which vendors run a recurring monthly allowance rather than a trial, what each allowance is worth in documents, whether a card is required, and what the first paid document costs on the other side. Covers LlamaParse, Azure Document Intelligence, Tabscanner, Veryfi, ParseRail (listed as running none), Invoice Data Extraction, Parseur, Eagle Doc, Google Document AI, AWS Textract, Nanonets, DocuClipper, Taggun, Mindee, Reducto and Lido: https://parserail.thecompound.tech/guides/free-invoice-receipt-ocr-api
- Benchmarks — field-level accuracy of every endpoint against its golden eval suite: https://parserail.thecompound.tech/benchmarks
- OpenAPI spec: https://parserail.thecompound.tech/openapi.json
- Dashboard (keys + billing): https://parserail.thecompound.tech/dashboard
- Studio: https://thecompound.tech