Have the agent come back later
Let the agent check back at a time or when something happens, start follow-up sessions now or on a later day, and set how often the agent may come back on its own.
Some work has to wait, such as a deploy this afternoon, a long CI run, or a check a week after a launch. Instead of keeping its sandbox running, the agent sets a time to come back, and Evergreen brings it back to the session then. You can also hand later work to a follow-up session that starts on the day you name.
When the agent comes back on its own
The agent sets a return when it has to wait for something. You can also ask for one in your own words, such as “check the deploy tomorrow morning”. The agent can come back:
- at a set time, up to 7 days out, or up to 30 days out when you asked for it
- when a pull request merges or closes, or when its checks finish
- when a pull request’s change reaches an environment
- when another session finishes or ships
- when a Datadog monitor changes state
While a return is set, a card above the message box says when the agent comes back and why. Run it now brings the agent back at once, and Cancel wake drops the return. Everyone who can open the session sees the card, and the people who can act on the session can use its buttons. While the agent is still at work, the card shrinks to one line saying when Evergreen tells the agent, with Cancel wake beside it.


A session that waits only on its own return has the status Scheduled, and its badge says when the agent is back. It’s listed with the running sessions, never under Waiting for you, since nothing waits on you. To see only those sessions, pick Scheduled in the sessions list’s Status filter, or search for is:scheduled.
The Coming up for you list on Standup gathers the returns and follow-up sessions set on your sessions.
Start a follow-up session
Type /follow-up and a task in the session’s message box to start a follow-up session from this one, such as /follow-up check the signup data. It starts at once, and you own it. It takes this session’s project, repositories, permission mode, visibility, and collaborators.
To start it on a later day instead, add by and the day: /follow-up check the signup data by friday. It starts at 9:00 AM on that day, in your own time zone. Evergreen reads days such as tomorrow, friday, next week, in 2 weeks, end of week, and Oct 12, up to 30 days out. Until then, the card above the message box names the session it starts and when, with Start it now and Cancel start.
If you send the same words again, Evergreen points you to the session they already started or holds for later, and doesn’t start a second one. The agent can start follow-up sessions too, and Models explains which model they run on.
How often the agent may come back
The agent may come back on its own only so many times in a row before someone writes to the session, and the count starts over when a person sends a message. The limit is 6 unless you change it.
To set your own for the sessions you own, open Settings › Defaults, find Wakes in a row under Sessions, and choose Change limit. You can pick from 1 to 50, and an empty field uses the default. A running session follows the new limit from its next return. Whatever the limit, a session comes back on its own at most 12 times a day.
Tip
Raise the limit when you leave a session to watch CI or a deploy, so it doesn’t stop after six returns. Each return is a turn of the agent, so it costs like one.
In the terminal
evergreen wake list <session>lists when the agent comes back to a session, why, and each return’s id.evergreen wake run <id> <session>brings the agent back now, andevergreen wake cancel <id> <session>cancels the return.evergreen wake list --minelists what’s coming up across your sessions.evergreen wake limitsprints your limit,evergreen wake limits --in-a-row 20sets it, and--in-a-row defaultgoes back to the default.
Last updated