All changes
v1.17.0

Idle breaks, a night session that closes when the morning starts, splitting a session, and more

Idle breaks cut the gaps out of an automatic day

When a commit or an agent opens your day, the session stays open and each commit is priced from the activity before it. A commute between commits on your phone and the first commit at your desk was billed as work. A new opt-in setting takes those gaps out. See Idle breaks.

  • Take idle breaks is under Backfill in your time tracking settings: Never (the default), Outside working hours, or Always.
  • Idle for longer than sets the gap that counts as a break, 15 minutes by default, from 5 to 240.
  • Count agent work as presence, off by default, lets agent work you were at keep a gap as work.
  • A gap longer than the threshold breaks from its start to the threshold before the next activity, so the work that resumes keeps that much lead-in. With Outside working hours, only the part of the gap outside your working hours, weekends included, breaks.
  • Breaks apply only to sessions opened automatically, never to one you clocked in by hand.
  • Prompts in interactive Claude Code sessions count as you being there. Clocktopus keeps only the time of each prompt, never its text, and ignores prompts from headless runs.
  • The Time page shows a break as Break and Resume. Click either to see why it was taken, and Undo break to merge the two sessions back and re-price the commit after it as if the break had never happened.

A session the night opened closes when the morning starts

A session opened between midnight and 06:00, by a night commit, a night agent session or work carried over midnight, stayed open into the morning. The first commit after you came back was priced across the night: a 00:40 commit and a 09:30 one recorded 8h50.

  • A session opened in the night now closes at the night's last activity, unless work ran on past the end of the night with no gap longer than an hour. The night runs from midnight to 06:00, or to the start of your working hours when that is earlier.
  • The morning then starts as any morning would: where an agent session clocks you in, or where the next commit would open a day. The 09:30 commit in the example records 30 minutes.
  • An agent session in the morning can now clock you in after a night session, instead of finding the day already open.
  • The close shows on the Time page as Break and Resume, and works whether idle breaks are on or off.
  • Sessions you clocked in by hand, sessions opened from 06:00 on, and commits that follow each other within the night are left as they were.

Record a clock-out and clock-in you missed

You can now split a session in two on the Time page when you forgot to clock out and back in.

  • Inside a session, each gap between entries offers Split session under Add entry.
  • The dialog fills in the previous entry's time as the clock-out and the next entry's time as the clock-in, and you can edit both to the second.
  • Entries stay unchanged, in whichever session their time falls in. A split with an entry's time inside it, times outside the session or in the future, and locked days are refused.
  • When the next entry starts before your clock-in, a checkbox offers to shorten it so it starts at the clock-in.

See what changed on the changelog

What changed in each release is now on this page, starting with the releases already shipped. The landing page's footer shows the version you're running and links to the entry that covers it.

Fixes & improvements

  • Time a headless Claude Code session, such as one run with claude -p or under an Agent SDK host, spends waiting for your answer to a question or a permission prompt no longer counts as agent runtime. Sessions already recorded keep their runtime.
  • The Work page's Tasks tab fits a phone screen: a long goal title is cut short instead of widening the card, and task rows drop chips as they narrow so they no longer overlap the key.
  • Dependencies in the app and the CLI are updated to clear known security advisories. This release ships CLI 0.9.0.