# Outputs (/outputs)


An answer in a transcript is not a deliverable. When the agent produces a
report, a spreadsheet, a chart, or a document, it writes the file into its
`output/` directory and Tidebreak publishes it as an **output**.

Outputs are scanned for after every command the agent runs, so a file appears
as soon as it is written — you do not have to ask for it.

## Finding them

- A strip above the transcript shows the chat's outputs. Toggle it with
  **Show outputs** / **Hide outputs**.
- The **Outputs** panel, opened from the activity chip in the chat header,
  lists all of them with **Name**, **Type**, **Size**, **Revisions**, and
  **Updated**, with search and a type filter.

Opening one shows a preview. Charts written as Plotly figures render as real
figures, not images: hover for values, drag to zoom, and use the toolbar to
zoom, reset, or save a PNG. Backgrounds, gridlines, and type follow the app's
theme in both light and dark — but only where the figure left them unsaid. A
chart is the agent's artifact, so anything it set explicitly, including the
series colors, is kept.

## Versions

The filename is the identity. When the agent writes the same filename again,
that is a new **revision** of the same output, not a second file. The
**Version history** popover on an output shows the revisions and lets you look
at earlier ones.

History is append-only. Restoring an older revision publishes it as a new
revision on top; it does not erase what came after.

## Exporting

**Save as…** on an output opens your operating system's save dialog and writes
the file wherever you choose. This is a real export — the file leaves the app
and is yours.

The agent can also propose writing an output into a
[connected folder](/connected-folders), which you approve from the
transcript.

## Deleting

**Delete** on an output removes it from the list, and the toast that follows
offers **Undo**. It is a soft delete.