Session tools

What you have at hand in a session, next to the chat itself.

The side panel New

The side panel is left of the chat. It has four tabs about the session on screen:

Tab	What it shows
Files	The file tree of the worktree.
Network	Which session started which, for the session on screen. It updates live. Click a row to open that session. See Cross-session workflows.
Graph	The graph card of the session, or Link to Graph… for a session without a link. See Graphs and agents.
Artifacts	The files and links the session produced. See Artifacts.

Open a tab:

Files: click the Files button in the title bar, or press ⇧⌘B.
Graph: click the Graph button, run Toggle Graph Panel in ⌘K, or press ⇧⌘G.
Network: choose Show Session Network in the session menu, or click the Started by banner or a Network link in the chat.
Artifacts: click the artifacts button next to the follow-ups button.

The panel keeps its width and remembers the last tab. Without an open session, the Network, Graph and Artifacts tabs stay visible but disabled. The panel then says why, for example "Open a session to see its network."

On a phone the side panel is a drawer from the left, with the same tabs.

The side panel on its Files tab shows the worktree's file tree, with the Files, Network and Graph tabs at the top.

The side panel left of the content, on the Files tab. This picture is from 1.5.2, before the Artifacts tab.

See the output

The pane next to the chat shows what the agent produces:

Markdown and text render in the pane.
Images open inline in the answer and in a zoom viewer (pinch, double-tap or scroll to zoom).
PDFs and video open in the pane.
HTML and SVG files open rendered in the file viewer. See HTML and SVG preview.
Web pages open in the embedded browser. An agent can open a page itself with the MCP tool open_in_browser.
Mermaid diagrams and Vega-Lite charts in a code block show as a picture. See below.
Every file path an agent writes in an answer is a link, including ~/, ../ and paths with spaces. Reveal in Finder is one click away. See File references.

Mermaid diagrams

A ```mermaid code block shows as a diagram in chat messages, plans and Markdown file previews. The diagram follows the light or dark theme. While a message streams, the block shows its code until the block is complete.

The toolbar of a diagram can:

show its source,
open it full screen, with zoom and pan,
copy its source, an SVG or a PNG.

Invalid Mermaid code shows the source with the error.

A chat message in the light theme shows a Mermaid flowchart and a sequence diagram, each with a small toolbar in the top-right corner.

Mermaid flowcharts and sequence diagrams render in place in a chat message. The toolbar is in the top-right corner.

A full-screen dialog shows a sequence diagram filling the window on a dark background.

Full screen opens the diagram in a dialog. Zoom with pinch, the mouse wheel or a double-click, and drag to pan.

Vega-Lite charts

A ```vega-lite code block shows as a chart. The block holds a Vega-Lite JSON spec with inline data. Bar, line, area, scatter and the other Vega-Lite marks work. The chart follows the light or dark theme.

The toolbar shows the source, opens the chart full screen and copies the source, an SVG or a PNG. An invalid spec shows the source and the error.

Charts never load data from the network. Put the data in the spec.

A line chart in light theme, followed by a code block that shows the error "Charts cannot load remote data or files. Put the data inline." above its source.

A chart in the light theme. Below it, a spec that asks for remote data is refused and shows the error above its source.

HTML and SVG preview

.html and .svg files that agents write open rendered in the file viewer. This works from tool calls, the file browser, file links and attached files. The viewer has three buttons:

Source shows the code. Preview goes back to the rendered page.
Reload loads the file again.
Open in Browser opens the file outside Splitkick.

A ```html or ```svg code block in the chat gets a Preview button. Nothing renders until you click it. A pasted .html or .svg file also gets a preview.

The preview is a sandbox:

HTML runs with no network, no app storage and no access to Splitkick.
SVG shows as an image, so scripts in an SVG never run.

The preview works in the app, in Web Access and on the phone. Open in Browser has no sandbox: there the page runs like any local file.

The file viewer shows report.html as a rendered dashboard with tiles and a bar chart, with Source, Reload, Open in Browser and Edit buttons in the header.

An HTML report that an agent wrote opens rendered in the file viewer, in a sandbox with no network.

A chat message with an html code block and, under it, the sandboxed preview of a click counter that shows 2 after two clicks.

The Preview button on an html or svg code block opens the page under the block, in the sandbox, only after a click.

File references

A file path in an answer opens the file it names:

One file found: the link opens it.
Several files found: a click opens a picker. It lists the files and where each one is.
No file found: there is no link. The path shows with a dotted line and a "?". Its tooltip says where Splitkick looked.

When the agent writes the file later in the turn, the link appears at the end of the turn.

Images

Right-click an image in a chat message and choose Copy image to copy the picture itself, not its path. The image viewer and the file viewer have a Copy image button too. When the copy fails, an error message says why.
Right-click an image in a chat message and choose Mention in chat to attach it to the composer of the same session, like a pasted image.
In a local session the image viewer has Reveal in Finder. In a remote session and in Web Access the file is on the other machine, so the viewer shows Open in browser instead: it opens the image that the other machine serves.

Right-click menu on an image in a chat response, with Copy image as the first item.

Right-click an image in a chat message and choose Copy image to copy the picture itself.

Image viewer in a local session with the buttons Reveal in Finder and Copy image.

In a local session the image viewer shows Reveal in Finder and Copy image.

Image viewer in a remote session with the buttons Open in browser and Copy image.

In a remote session the image viewer shows Open in browser instead of Reveal in Finder; it opens the image URL that the remote Splitkick serves.

Right-click a message in the chat:

Copy response copies the agent's answer.
Add to follow-ups puts the selected text, or the whole message, on the session's follow-up list.
Fork from here starts a new session from this point of the conversation; see Fork a session.
On an image: Reveal in Finder, Mention in chat and Copy image. Mention in chat attaches the image to the composer.

Grab an element from a web page

In the embedded browser you can send a part of a web page to the chat:

Click Grab DOM element in the browser toolbar.
Select an element on the page.
Choose Send to Splitkick Chat.

Splitkick adds the element to the message you are writing in the active session. Grab works in the desktop app only, not in Web Access.

Session tools

The side panel New