Automatically extract data from PDFs, invoices, contracts, and forms — classify, validate, and route to your systems in seconds. Zero manual data entry.
Documents enter via email attachment, file upload, or scan. The API accepts PDFs, images (PNG, JPEG, TIFF), and processes them through Claude's vision capabilities.
Claude identifies the document type — invoice, contract, receipt, proposal, form, ID document, or correspondence — with a confidence score and keyword fallback.
Type-specific extractors pull every field: line items, dates, parties, amounts, clauses, signatures. Dates normalize to ISO 8601, currencies to decimal, phones to E.164.
Math verification (do line items sum correctly?), required field checks, duplicate detection (same vendor + similar amount within 7 days), and low-confidence flagging.
Documents route to the right system based on type and validation results. Issues, duplicates, or high-value items automatically escalate to human review.
Original documents are stored as-is in S3 alongside the extracted JSON. Full audit trail — every field traceable back to its source page and location.
Slack alerts with document summary, routing decision, and any flags. Calendar events auto-created for due dates, renewal deadlines, and notice periods.
Each document type follows its own extraction and routing logic
Vendor matching, PO reconciliation, math verification, and auto-bill creation for amounts under threshold.
Clause extraction, key date computation (expiration, renewal notice), unusual term flagging, signature tracking.
Auto-categorization (meals, travel, supplies), tax deduction tracking, threshold-based approval routing.
Key-value pair extraction from any form type, field normalization, missing required field detection.