Use cases
Seven moments where the work stops waiting.
A review between meetings, a crash triaged overnight, a question answered in a thread. Four are for the developer with a phone, three for the team in Slack — pick the one that sounds like your week. Each one also says plainly where it stops.
For the developer
The agent keeps working. The review comes to you.
Four moments where the work would usually wait for you at the desk. Your computer keeps the code and runs the agent; you still read the diff, from wherever you are.
See it for developers →For the developer
A pull request review that ran the tests, from your phone
A useful review means running the code. Your computer checks out the branch, runs the setup script, builds, and tests — and one review lands on GitHub with its comments anchored to real changed lines.
gh CLI · disposable worktree · host-owned postingFor the developer
Three branches, three agents, one phone
A refactor, a bug fix, and a dependency bump, each in its own worktree and branch so their files stay apart. One Activity screen shows which of them needs you.
isolated worktrees · snapshotted base · cross-host ActivityFor the developer
Let your computer work overnight
The full suite is too slow to run while you work. Put it on a cron line: every fire records an outcome, and a worktree stays only when there is something to review in the morning.
five-field cron · recipes · durable historyFor the developer
Start the small fix from your tracker, on your phone
The small fix usually waits for the desk. A work-item recipe reads the board your agent already reaches — you name the tracker in the prompt — and one item becomes a workspace on your phone.
work-item recipes · Act-mode MCP · chat-first workspaceFor the team
The channel asks. The code answers.
Three moments that used to cost an engineer’s time: a question, a 2 a.m. crash, the morning report. On a computer the team owns, each one ends in a real agent session on real code, and the engineers keep coding.
See it for teams →For the team
A crash at 02:14, triaged before anyone wakes up
The forty minutes spent finding which code a crash belongs to happen overnight. The reporter calls a URL, your verifier admits it, and a triage recipe reads the stack against the real checkout and files the story.
trigger URL · verifier · Act-mode recipe · the agent’s own trackerFor the team
Ask the code in Slack, not the engineer in the meeting
The answer is in the code, and the person who knows is busy. Ask in a bound channel; the answer arrives in the thread, footed with the project, commit, branch, and mode it was read under.
bound channel · Act-mode inquiry · answer footerFor the team
The morning report is already written
Nobody has time to write the daily summary. One recipe on one cron line posts it into the channel through the Slack notifier — and it can say “since last time” because it knows when it last succeeded.
cron · recipe · {{last_success_at}} · Slack notifierAll seven run on a computer you own.
Install the host once. Your phone and your team’s Slack channel are two ways to reach it.