LightOn

LightOn provides three main endpoints: /parse converts documents (including scans, tables, handwriting, and multi-column layouts) into structured Markdown, supporting 20+ languages natively. /extract allows users to define a JSON schema and extract specific fields (like invoice numbers or contract clauses) from documents, returning structured JSON. /search performs hybrid retrieval using dense, sparse, and late-interaction signals, and returns results with citations at the chunk level. The platform is LLM-agnostic, MCP-native, and offers workspaces with ACLs at the chunk level for fine-grained access control.

LightOn is designed for AI agents that need to handle messy real-world inputs, such as raw PDFs, garbled tables, and off-domain queries. It supports conversational search and Q&A, massive multimodal RAG (ingesting images, diagrams, and handwritten notes), tool chaining for complex multi-step tasks, and the creation of custom specialized agents for different teams. It also enables universal data synchronization from external sources like SharePoint, Drive, and Confluence, with strict governance and ACLs to ensure data security and compliance.

LightOn offers a free starter tier with usage-based pricing (PAYG) and no commitment. The Business plan costs €149 per month plus PAYG, and includes an annual license and business SLAs. For large organizations and the public sector, there is an Enterprise plan with custom pricing, tailored deployment, enterprise SLAs, and white-glove support. Specific per-page and per-query prices are listed: /parse costs €0.002 per page, /extract costs €0.004 per page, and /search costs €0.006 per query.