Watch Your Agent Work in Slack

Mentioning an agent in Slack used to mean staring at a ⏳ reaction until the answer showed up. A ten-minute task and a stuck one looked the same. Now you can follow along.

What shows up in the thread

• A working line. As soon as the agent starts, Slack shows it's working under the thread. The line clears when the answer posts.
• A live checklist. A "Working on it" reply lists each step as it happens and is edited in place. When the turn ends it turns into a summary like "Worked for 2 min · 12 steps". The answer still arrives as its own message, formatted the same as before.
• Pull request cards. When the agent opens or pushes to a pull request, the thread gets a card with the PR title, repository, number and diff size, plus a link to open it. The checklist of the turn that opened the PR adds the diff size to its title too.

Readable steps

The checklist is meant to be skimmed. Reads, searches and web lookups fold into one "Looked around" step. A step that repeats shares a row with a count, as in Ran go test 3 times (2 failed). Claude agents describe each command in a few words ("Check the migration ran"), and that description is what you see.

Raw commands never appear. Steps show file names, program names and the agent's short description, and anything that looks like a token or password is masked as **** before it reaches Slack.

Where it applies

It's on for every organization, for tasks a person starts in Slack. Tasks started by another bot or by an automation look exactly as before, so a workflow that reads the agent's first reply still gets the answer. Discord is unchanged.