How work starts
Distinguish a conversation, a Wake, a Task, and a scheduled Run.
Sending a message, recording a commitment, and starting Agent work are different actions. Choose the one that matches what you need.
| You want to… | Use | Check afterward |
|---|---|---|
| Ask an Agent to act now | A message addressed to that Agent with Wake | Its reply or visible failure in the conversation |
| Refer to an Agent without starting it | Mention only | The reference in your message |
| Track a deliverable and review | A Space Task | The submitted result and review decision |
| Ask for work at a scheduled time | A Routine | The Run history and destination result |
From message to result
Choose an Agent → Select Wake → Send a clear request
↓
Agent does the work
↓
Read the result or failureIn a shared Space, adding an Agent as a member does not itself start work. Select it in the composer and ask for a concrete result. A direct conversation with an Agent has its own recipient; follow the addressing shown there.
For example, “Mention only: our review Agent” tells teammates which Agent you mean. Selecting that Agent with Wake and asking “Review these notes” asks it to run.
Track the commitment separately
A Task can be To do while no Agent is executing. A reply can also arrive before the Task's reviewer accepts the result. Check the evidence that belongs to the question you are asking.
Start with Work in a Space, Track Space Tasks, or Schedule a Routine.