Wake-ups & fan-out
Work that does not fit one turn: come back later, or do the same thing for every item in a list.
| Tool | What it does |
|---|---|
schedule_followup | Schedules the agent to run again later with a follow-up prompt โ a reminder, a deferred check, "circle back in three days". |
map_over | Runs one child agent per item of a list or CSV with the same prompt ({item} is the placeholder), in parallel, then collects the results into one artifact. |
schedule_followup
Schedules the agent to run again later with a follow-up prompt โ a reminder, a deferred check, "circle back in three days".
How an agent gets it. Added automatically on chat and automation runs.
- Time is ISO 8601 or relative (
+2h,+3d). - Bounded by the organisation's
wakeuppolicy: how far ahead, and how many pending per person and per agent. Pending wake-ups appear under Automate.
Arguments
| Argument | Type | Required | Meaning |
|---|---|---|---|
when | string | yes | ISO 8601 datetime or a +Nh/+Nd offset. |
prompt | string | yes | What to work on when you wake up. |
map_over
Runs one child agent per item of a list or CSV with the same prompt ({item} is the placeholder), in parallel, then collects the results into one artifact.
How an agent gets it. Added automatically on top-level chat and automation runs โ never on a child run, so fan-out is one level deep.
- Bounded by an item-count cap and a cost cap over the whole family of runs; one failing item does not fail the others.
target_agentruns a different agent per item. See Orchestration & autonomy.
Arguments
| Argument | Type | Required | Meaning |
|---|---|---|---|
over | array | string | yes | A JSON array of items, or a CSV/newline-delimited string. |
prompt_template | string | yes | Prompt for each child; use {item} as the placeholder. |
target_agent | string | Optional slug of a different agent to run per item. |