What the agent can create
Files, exports, notes, citations, and images the agent produces from your chat, and how to ask for each.
Beyond answering questions, the agent can turn its work into something you keep: a file to download, a note in your library, a formatted reference list, or an image. Ask in plain language, or type @ and pick a tool from the Tools group.
Files
The agent can create Word documents, PowerPoint presentations, spreadsheets, and charts as downloadable files, on every plan. It can build a spreadsheet from data in your documents, edit an existing one, and query it for analysis. Charts and graphs come from the data in your sources.
Build a spreadsheet comparing sample size, intervention, and primary outcome across the trials in this folder, then chart the effect sizes.Exports
Anything in the conversation can be exported as a DOCX, PDF, BibTeX, or CSV file. Say which content you want and the format.
Export the summary table above as a CSV, and the papers it cites as BibTeX.To export a note you have written instead, use the note menu. See Write notes.
Notes
The agent creates a note only when you ask it to, so answers stay in the chat unless you want them saved. It can also edit a note you already have, such as adding a section or rewriting a paragraph. Use @Create note or @Edit note to make the request explicit.
@Create note Save this comparison as a note called "Methods overview".Citations
The agent formats citations for documents in your library in APA, MLA, Chicago, Harvard, IEEE, CSE, ACS, Vancouver, AMA, or ASA. Use @Create citation, or ask for a reference list in the style you need.
@Create citation Format every paper in my thesis folder in Vancouver style.These are separate from the inline citations in every answer, which follow your citation format setting. See How citations work.
Images
The agent can generate a new image or edit one you attach, which is useful for diagrams, illustrations, and quick visual edits. Images appear in the chat.
@Create image Draw a simple diagram of the study design: two arms, randomization, 12-week follow-up.