Skip to content
DocDesk

Document workspace

Chat with your documents.

Use DocDesk to summarize a report, outline a presentation or check a spreadsheet. Attach up to eight files to one conversation, review the sources and download an editable result.

Worked example
Stock count.csvCSV
ItemUnits
Notebooks17
Folders25
Show table statistics for this file.
Total units42

Inspired by Anthropic’s Claude Skills

From source files to editable results.

DocDesk takes inspiration from Anthropic’s publicly shared Claude Skills document workflows and independently implements a connected process: read your files, organize the content, then export a result you can keep editing.

  1. Bring your files together

    Attach up to eight documents, PDFs, spreadsheets or presentations in one conversation. Older DOC and PPT files convert automatically, saving you a separate conversion step.

  2. Check the source as you work

    Use source excerpts and references to check summaries and drafts against your files. Keep the original material close as you review and refine the result.

  3. Keep working after download

    Turn a summary, table or slide outline into an editable Word, Excel or PowerPoint file. Continue refining it in your usual tools and share it when it is ready.

The studios

What are you working on?

DOCX · DOC · MD · TXT

Summarize a document with sources

Turn a report or meeting notes into an editable summary. Follow each source reference back to its document before sharing the draft.

Before you upload

See formats and limits
Which files can I use?
Upload DOCX, PDF, PPTX, XLSX, CSV, Markdown and plain text files up to 20 MB each. Older DOC and PPT files convert automatically up to 10 MB each. Save old XLS files as XLSX first.
Can DocDesk read a scanned PDF?
DocDesk extracts the text layer of a PDF. Run optical character recognition (OCR) on image-only scans before uploading. Unlock password-protected PDFs first.
Which tasks require an AI model?
Source extracts, keyword searches, template drafts and table calculations work without a model. Free-form rewriting and semantic follow-up questions require a configured model. The studio shows the current mode.