Koncile
Koncile is a document-processing platform for teams that need structured data from invoices, identity documents, bank statements, and other business records. Its described workflow combines image preparation, OCR, field extraction, and checks against information elsewhere in a file or connected database. The useful distinction is between reading a document and deciding whether its contents require closer review.
Extract fields into usable records
Koncile describes cleaning document images before reconstructing text, tables, columns, and paragraphs. It then converts that information into structured fields rather than leaving a team with a searchable scan alone. Koncile says its outputs include JSON, CSV, XLSX, and Excel, with a confidence score for each field. That gives an integration a way to distinguish an extracted value from a value that needs confirmation.
For invoices, the homepage describes line-item extraction alongside totals and tax checks. For a dossier containing several document types, Koncile says it separates the file, classifies each document, extracts the relevant fields, and cross-checks them against a customer database. These are publisher-described capabilities, not results from a Franklin accuracy test.
Review inconsistencies and fraud indicators
The platform describes checks for mismatched information, missing documents, and inconsistent calculations. Its fraud features look at signals such as PDF structure, fonts, metadata, and image characteristics, with document-specific context contributing to the assessment. A flagged file or score should be treated as a reason to investigate, not proof that someone submitted a fraudulent document.
Koncile also describes no-code alerts for totals, thresholds, dates, and other business rules. A finance team could therefore evaluate whether its own acceptance criteria can be expressed in the workflow, instead of assuming one generic document score answers every operational question.
Connect extraction to existing systems
The publisher lists a REST API, Python SDK, structured JSON responses, and webhooks for asynchronous workflows. It also describes an MCP OCR server for agent access. The practical evaluation is whether the extracted fields, confidence information, and review flags fit the system that will consume them.
Before using it for consequential decisions, test representative documents and inspect incorrect or uncertain outputs. Confirm integration behavior and review responsibilities separately from the publisher's performance and security claims; the captured homepage does not establish an independently verified accuracy rate for a particular document collection.