Runtime

In workspaces where agents can work on their own computers, each job shows on the card or in the chat as a single status line (see Work on the agent's own computer). The Runtime page shows all of that work together: what is running now, what waits in line, what finished today and in the last seven days, how it ended and how many credits it used.

The page is the Settings › Runtime tab. Only workspace admins see the tab, and it appears only in workspaces where agents working on their own computers is turned on.

#What the page shows

• Summary. One box for today and one for the last 7 days: jobs finished, average time, highest memory use and credits spent. Below them, how the jobs ended, with counts.
• Running now. Each job is one line: which card or flow, which agent, who it works for, its status and how long it has been running. Click the card or flow name to open it.
• Waiting in line. Work waiting for a free computer. It starts on its own when one frees up; this part shows only while something is waiting.
• Finished work. A paged list for today or the last 7 days: when it ended, which work, which agent, how it ended, time, highest memory and credits.

Chats are not shown by name, only as "Chat": a chat belongs to the person who opened it. A card in a closed space you are not in shows by its number only, without its name or a link.

#How work ends

On the page	Meaning
Finished	The agent finished its work.
Closed after nobody wrote	The chat or question went unanswered for a while, or the idle computer made room for work waiting in line. The next message continues where it stopped.
Stopped	Someone stopped the work.
Time limit reached	The work reached the longest time allowed, or the computer sat idle for a long time. Long work that is still going continues on a new computer.
Memory limit exceeded	The work did not fit in the computer's memory. The next job of the same card opens with more memory.
Safety limit reached	The work reached the spending limit that stops a runaway loop.
Model changed	The chat's next message asked for another model; the new computer opened with it.
Tools or connections changed	The agent's tools or connections changed; the next job opened with the new permissions.
Work moved to another space	The card or chat left a closed space; the work starts again with an empty folder so its files do not reach the new space.
Card or chat deleted	What the work belonged to was deleted.
Ended with an error	The work stopped unexpectedly; the card's error note says why.

#Stopping work

Every running job has a stop menu on its line. It is the same menu as on the card and in the chat:

• Stop after this turn. The agent finishes the step it is on and does not start another.
• Emergency brake. The agent stops at once and the work may be left half done. The files written so far are kept; if the work starts again it continues from there.

A stopped card goes back to review. Every stop is written to the Inbox › Audit tab with who stopped it.