Koncile

Market position

Koncile is described by an industry directory as a Paris-based intelligent document processing platform competing directly with Amazon Textract, Mindee, Nanonets and Klippa. It reports more than 10,000 users and over 100 clients. Third-party estimates place headcount at 1-10 and estimated annual revenue near USD 513,000, indicating an early-stage position relative to larger IDP vendors.

Koncile positions itself against traditional OCR tools by adding an LLM comprehension layer that interprets context rather than transcribing text, and by allowing users to define extraction rules, transform data and run analysis inside the interface. The company also emphasises French design and hosting, GDPR compliance, SOC 2 certification, advanced encryption and a commitment not to train models on customer data. Its own competitive comparison claims it is the only tool combining a no-code interface with a complete API and structured multi-format export; a third-party directory notes this claim is self-reported and not independently verified, as are the published accuracy benchmarks. The MCP OCR server, exposing 24 tools for AI agents, differentiates it as infrastructure for agentic document workflows.

Technology

The platform combines an OCR engine with large language models in a three-stage pipeline: pre-processing (skew correction, resolution normalisation), OCR conversion of image to text including printed content, handwriting, tables and non-standard layouts, and LLM analysis that structures output and assigns per-field confidence scores. Handwriting recognition uses LLM-based and deep-learning models within the same pipeline. Capabilities include automatic document classification, page splitting, table extraction, data categorisation, database matching and rule-based anomaly alerts. Outputs are available in JSON, XML and CSV; access is via a no-code web app, REST API, SDK libraries, a Bubble plugin and an MCP server. Security posture includes GDPR compliance, SOC 2 certification, advanced encryption, French development and no model training on customer data. An external profile tags the technology stack as big data, artificial intelligence and machine learning with a B2B business model.

Go-to-market

Go-to-market combines product-led self-service (free trial credits, no-code template builder, three-minute test, public documentation and template library) with sales-led demo booking and content marketing through a blog covering analyses, feature posts, comparatives, a glossary, news and tutorials, plus OCR benchmark and comparison pages. Developer distribution runs through a REST API, SDK libraries, a Bubble plugin and an MCP server. Third-party reporting describes a deliberate move from horizontal OCR into the procurement vertical, with the ADRA'Up award providing visibility inside the French procurement community.

Finance, accounting, procurement, compliance/KYC, audit and logistics teams in businesses of varying size, plus developers integrating extraction into their own products. Named reference customers and testimonial sources include Grainz, Sightness, Demathieu Bard, SpendHQ, Béal, Cosma Experts, RewardPulse, Bilfinger, Place des Energies, Mark Young Construction and Nona. Documented user personas include accountants extracting pre-tax invoice amounts, compliance officers running KYC checks, procurement teams building databases from invoice line items, warehouse managers matching delivery notes to invoices, and healthcare organisations processing medical forms and handwritten prescriptions.

Geography

Headquarters are in Paris, France, with the company's website footer listing 96 bis Boulevard Raspail, Paris 75006, and a data aggregator listing 92 Boulevard Raspail, Paris 75006. The company describes itself as a French startup with a product designed and developed in France. Its site is published in French and English, and testimonials reference customers in France as well as at least one US-named construction company.

History

Sources give conflicting founding information: the company's own ecosystem listings and a data aggregator state 2023, while an industry directory states 2022 and names three founders (Jules R., Tristan Thommen and Raphael Balbous). Early backing is attributed to the Agoranov incubator and Entrepreneurs First. By 2026 the company reported more than 10,000 users and over 100 clients, published a public documentation site and API, won the ADRA'Up Startup of the Year award in January 2026, and launched an MCP OCR server in March 2026.

Risks & controversies

Public sources conflict on basic facts: founding year is given as 2022 in one directory and 2023 in others, founder counts differ (two versus three), and an autocreated ecosystem profile lists an implausible creation date of 2000 and zero funding raised. Performance claims — 98.6% OCR accuracy, 99% accuracy on simple fields, 95-96% on invoice line items, 800 hours saved, up to 20% purchase savings — are vendor-reported and, per a third-party directory, lack independent verification, as does the 'only tool' competitive claim. Revenue, valuation and funding figures from data aggregators are algorithmic estimates and internally inconsistent (one page simultaneously lists 'No funding' and a 2024 funding round with no amount).