BE AI runs. A row is one investigation on a computer you were given. There is no token column and no cost column.
Open BE AI runs. One row: Accounts server, alert:disk_high, the question “Accounts server disk / is at 92%. What is using the space, and is it safe to clean?”, status done, result known_growth. The result column starts the summary: the volume is 92% because of SQL backups in /var/opt/mssql/backup. There is no tokens column and no cost column. The count says 1 of 1 runs, and 4 tool calls. Click the row. This shot has not.
The detail names the recipe disk_high, the trigger alert:disk_high, that the system started it, and that it is done after 4 tool calls. There is still no token count. The classification is known_growth. The summary says the volume is 92% because of SQL backups in /var/opt/mssql/backup. Safe clean-up would reclaim about 12 GB of rotated dumps older than 14 days. A person should approve that. The next step is Approve deleting rotated SQL dumps older than 14 days. Three proposals sit under that, as status chips. Measure /var/log is executed. Vacuum the system journal is failed. List the largest folders on / is proposed. The reason column is why each command was suggested. The failed chip does not repeat the sentence from the queue, allow_shell is off on this machine. The transcript is empty. Opening the row does not approve a proposal and does not start another run.
The list before the click is See BE AI runs on your machines. A word in the question is Search BE AI runs on your machines. One machine is Show BE AI runs for one of your machines. A recent window is Show BE AI runs from a recent window. The waiting command is A proposal you cannot run. What your provider sees, including the token cost, is Read one BE AI investigation. Usage is not in this nav.