Extend

A document processing API that parses, extracts, and splits unstructured documents with high accuracy, solving the problem of converting messy documents into usable context for AI agents and pipelines. It is built for developers and engineering teams at AI companies and enterprises who need reliable document processing for production workflows. The service is delivered as a cloud API with multiple performance modes, a toolkit including confidence scoring and a composer agent for schema optimization, and a studio interface for evaluation.

Key features

• Layout detection for tables, checkboxes, images, handwriting, signatures
• Specialized vision models for each document element
• Multiple performance modes: speed, cost, accuracy
• Handles 25 file types, 100+ languages, 4 chunking strategies
• Confidence scoring to flag uncertainty
• Fast mode for low-latency or cost-optimized processing
• Composer agent for automatic schema refinement
• Document workflows for end-to-end orchestration
• Studio and evals for schema iteration and regression catching