CiteLLM

CiteLLM is a verification layer for AI answers drawn from documents. It offers a ChatGPT and Claude plugin (Check) to verify claims and trace answers to highlighted passages in the original PDF, a document search and question answering tool (Ask) that returns passage-level citations across a document collection, and a structured extraction tool (Extract) that pulls defined fields into structured data with a source for every value, with an API available. Results can be reviewed and approved or flagged, and an evidence report can be exported. The page states that files are not used for training, are deleted after processing by default unless kept, and can be self-hosted with Docker. Example use cases shown include covenant compliance checks, contract questions, lease abstraction, brief cite checks, insurance claim coverage questions, and invoice extraction.

Key features

• Checks claims in ChatGPT or Claude via plugin, marking each as supported, contradicted or missing
• Highlighted source passages in the original PDF for each answer
• Document search and question answering across a document collection with passage-level citations
• Structured extraction of defined fields with a source and confidence score for each value
• API available for extraction results
• Approve or flag each result and export an evidence report or highlighted PDFs
• No training on customer documents
• Files deleted after processing by default unless the user chooses to keep them
• Self-hosting via Docker

Best suited for

• Finance teams checking covenant compliance and credit agreements
• Legal teams reviewing contracts, briefs and depositions
• Real estate teams abstracting leases
• Insurance teams answering coverage questions from policy wording
• Audit and compliance teams extracting invoice data
• Regulated teams needing to show where figures came from

Pricing

Not stated on the website