Skip to content

Activity and notifications

Alfred does things when you aren’t in a chat with him: jobs run, reflexes fire, reminders go out, and he updates his memory. Three places on butler.now tell you about it.

AnswersWhere
NowWhat is Alfred doing right now, and does he need me?Alfred’s face at the top left of every page
NotificationsWhat has Alfred told me?Open the full feed, from Alfred’s face
ActivityWhat has Alfred done on his own?Activity, in the Chat panel of the Dashboard

The Butler logo at the top left of the site is Alfred. When he is working on something, such as a reply, a job or a background agent, he says so beside his face, counts you down to a meeting, and asks there when he needs an answer. A count on his face marks what he told you that you haven’t seen.

Click him to open what he is on: everything running right now, any question or approval waiting on you (which you can answer there), and what he told you recently. Mark all read clears the count.

Your notifications are everything Alfred told you outside a chat reply, newest first: a job’s question or approval, an update on a job, a message you asked him to send you, a reflex that stopped working. Each one is what he sent you, as you received it on Discord or anywhere else. Ones that still need an answer say so, and you answer them from the feed.

What counts as unread is everything since you last looked: opening the feed, or Mark all read, moves that line forward. The count shows up to 99, then “99+”.

The activity log is what Alfred did without being asked in that moment, newest first, one line per thing with the details under it. Filter it by kind:

KindWhat it records
JudgeThe decisions a job’s autonomy made for it: a call it approved, refused or asked you about, and why.
BrainWhat Alfred wrote to his memory, and from which conversation. See Brain.
OutreachQuestions and approvals a job closed for you: withdrawn when it changed its plan, closed when it finished, or answered by something you had already said.
AgentsEach run of a background agent: what woke it and what it did.
RemindersReminders that fired, and where they were sent.
ReflexesWhat a reflex did, and a reflex paused because it kept failing. See Automation.

Under All, the log also holds:

  • Discord messages Alfred judged were for him.
  • What was submitted on your places, and the buttons pressed on them.
  • What your watches did: fired, paused, ended or cancelled.
  • Changes to your projects.
  • Big moves and deletes in your drive, every delete for good Alfred asked you to approve and how it went, and files visitors sent.
  • Your secrets and webhooks being made, changed or removed. Never a key itself.
  • Settings Alfred changed because you asked him to.
  • Writes your datasets refused.
  • Every outward thing a job, a reflex or a background agent did with nobody watching: a message sent, a card changed, a key used. One that overwrote a card says which version a restore puts back.

An entry that belongs to a job opens the job, and one that names a card says which. Entries are kept for 90 days, and a watch’s for 30.

The activity log is a record, not a list of things to answer; anything that needs you also arrives in your notifications.