Skriptr: AI writing workspace that cites only from your uploaded sources

Skriptr launched on Product Hunt on August 17, 2026 as an AI research and writing platform built around a specific constraint: the AI only draws on documents in the user’s personal library. It does not generate content from its training data or produce citations from memory. Every claim it surfaces links to a specific page in an uploaded source.

This design directly addresses the main objection to AI in research-intensive writing — that AI systems produce plausible-sounding citations to sources that do not exist, or accurately represent sources the writer has not read. Skriptr avoids this by functioning as an AI layer over the user’s own uploaded materials.

The platform offers several AI modes suited to different stages of research and writing. Learning Companion explains concepts from the uploaded documents. Literature Review maps the field as represented across the uploaded sources. Devil’s Advocate argues counterpoints to a position using evidence from the library. Each mode produces output that appears as tracked changes — every AI suggestion is a redline that the writer explicitly accepts or rejects, maintaining authorial control over the final document.

Collaboration features allow shared libraries and real-time document editing. The platform accepts PDFs, Word documents, web pages, images, recordings, and notes, and exports to Word, PDF, or Markdown.

The free plan supports five source uploads with limited AI usage. Pro (USD 10 per month) increases to 100 sources with audio transcription. Ultra (USD 30 per month) offers maximum usage and early feature access. User documents and drafts are not used for model training.

The product is aimed primarily at university students — current institutional coverage includes the London School of Economics, Boston University, and several Scandinavian universities — but the source-grounding model applies to any research-intensive writing where citation accuracy matters.