Definition fields
A scheduled agent is defined by four fields, set from Scheduled Agents in the Agents panel. One of them, accounting basis, determines the figures every run returns. The entity a run reports on comes from the wording of the instruction. Scheduled agents are private to the person who creates them. Creating, editing, pausing, resuming, running, deleting, and rolling back an agent are restricted to organization admins, and an admin sees only their own agents in the list. There is no way to share an agent or its runs with a colleague. Someone else who needs the same report creates an agent of their own. The table below lists every field on the agent definition.
There is no entity field. An instruction that names no company runs across every company the owner can reach and returns consolidated figures, the same as a question typed into the composer. Agents created while the definition still carried a company selector keep the company they were created with, and every run of those agents stays scoped to it.
An organization can hold 50 scheduled agents at a time.
There is no per-agent connector selection. An agent reaches the same connectors as the person who owns it, which are the ones that person has personally connected. See how to connect an MCP connector for why connections are per-user rather than per-organization.
Cadence options
Cadence is either a repeating interval or a calendar schedule, and the picker offers five modes. Every calendar mode carries its own timezone.
There is no separate daily mode. For a once-a-day agent, set an interval of one day, or use Weekly with every weekday selected if you want it to skip weekends.
Times resolve against the timezone stored on the schedule. Any IANA timezone is accepted, such as
America/New_York. A schedule saved without one resolves against UTC.
The minute you choose is adjusted by up to four minutes when the schedule is saved. DualEntry spreads scheduled runs across a five-minute window so that agents created on the same round hour do not all fire at once. An agent set for 09:00 may therefore run at 09:03.
Where results appear
Every run of a scheduled agent gets its own chat session, and the runs list in the agent’s drawer is where you find them. Open the agent from Scheduled in the Agents panel or from the Scheduled Agents page to reach its drawer. The drawer carries the definition, the current status, the last error, and the most recent runs. Each run row shows when the run started and its status. A finished run opens as an ordinary chat: the instruction as the first message, the agent’s response beneath it, and a composer you can reply in. Open latest run in the drawer footer opens the agent’s most recent run when it reads Completed, Failed, or Waiting on you. It is disabled while that run is still running. It does not skip back to an earlier run. A run that is still running cannot be opened. The sidebar entry, Open latest run, the running row in the list, and a direct link to the run all fall back to the drawer. The controls are disabled, and a tooltip explains that the run is still in progress. Once the run finishes, the same controls open it. There is no email digest, no Slack delivery, and no notification of any kind. The agent’s entry in the Agents panel shows a badge counting the runs you have not opened yet. The line under its name reads Ready to view when the latest run has finished, so the panel tells you there is something to read without telling you what it says. An agent whose output nobody opens produces nothing that anyone reads. Runs are private to the agent’s owner. Nobody else can open a run chat, and a run cannot be shared with the organization. If a colleague needs to act on the output, pass it on from the run, or have them create an agent of their own. Runs stay out of Recents until you reply in one. Recents lists conversations you started, and a daily agent would bury them. A run you have replied in has become your conversation, so it appears there from that point on. A run you only read stays findable from the agent’s runs list. Deleting an agent removes it and its runs list from the panel and closes every run chat. A closed chat is read-only and no longer appears in chat lists; it is not deleted. Pause the agent instead when you want the schedule stopped and the runs list kept where it is.Run behavior
A scheduled agent runs unattended, which changes what its permissions, its scope, and its failures mean compared to the same instruction typed into the composer. An agent runs under the permissions of the user who owns it. Results reflect that person’s role scope, which matters when the owner changes teams or leaves. Review agent ownership when roles change. Accounting basis comes from the agent definition rather than from a selector someone sets at run time. For agents created without a company, entity scope comes from the instruction. An instruction that names no company returns consolidated figures on every run, with no one in the composer to catch it. Agents created while the definition still carried a company selector keep that stored company on every run, whatever the instruction names. Rewording the instruction does not move such an agent to a different entity. Each run is written to the audit trail with the same detail as an interactive conversation, attributed to the owner and tagged as a scheduled run, so an unattended result is as traceable afterwards as one you asked for yourself. Every run carries one of four statuses, shown on its row in the runs list and reflected under the agent’s name in the Agents panel. The table below lists them.
A run left in Running because the worker behind it died is marked Failed after about five minutes, with an error noting that the turn was abandoned. A schedule therefore cannot stay stuck behind a dead run.
If the previous run is still going when the next tick fires, the tick is skipped rather than queued. A skipped tick produces no run at all, so it shows up as a gap in the runs list rather than as a row with a status.
A failed run increments a failure counter on the agent. Three consecutive failures move the agent to an errored state and stop the schedule, which keeps a broken instruction from running every day indefinitely. The owner is not notified when this happens beyond the Run failed line under the agent’s name.
Pausing an agent stops the schedule and keeps both the definition and the runs list. Resuming restarts it on the same cadence and clears the failure counter. The drawer carries the current status, the last error, and the version history of the definition, so a result you cannot reproduce interactively can be traced to the wording it ran under.
Run Now
Run Now starts a run immediately, outside the cadence. The button sits on each agent card and in the drawer footer, and it is restricted to organization admins, the same gate as Edit. A manual run takes the same path as a scheduled tick: the same instruction, the owner’s permissions, the same read-only tool set, and its own chat session at the top of the runs list. The difference is in the health counters. A manual run that fails records Last Error on the agent but does not count toward the three consecutive failures that stop the schedule. Run Now works on a paused agent, so you can check an instruction without resuming the schedule. An errored agent has to be resumed first. Two conditions block a manual run:- A run is already in progress. The latest run is still running, or is waiting on you. Wait for it to finish, or open it and respond.
- The agent just ran. Run Now is refused for one minute after a run starts, so a double click cannot stack two runs. Try again after the minute passes.