@doconda/editor). It comes with an assistant: the user clicks ✨, types what they want (“make it more formal”, “add a
row with the March total”, “a closing slide”) and the assistant makes the changes in the document itself while it
answers. The user accepts or discards them in one go, and the file is still a normal .docx, .xlsx or .pptx.
What it can do
The assistant reads the whole document, so it understands requests like “sum up the sales table in one paragraph” or
“colour the expenses over budget in red”.
How it fits in
- A side panel (the default), a floating box or your own component, with
assistant={{ layout, render }}. - Example requests with
suggestions, and your own field text withplaceholder. - From your code:
editor.current.assistant.ask("Translate the document into Spanish"), for example from your own button. - No review step, with
review: false, if you’d rather keep changes straight away. - Remove it:
assistant={false}.
Privacy and price
Opening, editing and saving happen in the browser. Only when the user asks the assistant for something do the document text and the request travel to Doconda to make the change, with models that run in the EU, and they are not stored. Withassistant={false} nothing leaves the browser.
Every editor plan includes 2,000 edits a month per organization; after that, each edit costs €0.01 from the
balance. A request that changes nothing or fails doesn’t count. See license and domains.
Compared with other options
ONLYOFFICE, Collabora Online and Syncfusion also have AI features, but they ask you to connect your own key from a model provider. In Doconda the assistant is included and makes reviewable changes inside the document. See the comparisons with ONLYOFFICE, Collabora and Syncfusion.Frequently asked questions
Does it work with PDF?
Does it work with PDF?
The assistant edits Word, Excel and PowerPoint. For PDF the editor fills in forms, annotates, comments and
organizes pages, without the assistant.
Can it undo what it did?
Can it undo what it did?
Yes. Changes stay pending until the user accepts them; “Discard” removes them all, and Ctrl/Cmd+Z undoes them too.
How is this different from editing with the API?
How is this different from editing with the API?
With the API, your agent changes the document on your server (edit a Word file with AI).
With the assistant, the user asks from the editor and sees the changes before keeping them.