Automatic time tracker: what it records and how it writes your timesheet
An automatic time tracker records your work in the background and drafts the timesheet. What it captures, how activity becomes entries, and how Planim does it.
Contents
An automatic time tracker is an app that records what you work on in the background and turns that record into time entries, with no timer to start or stop. Planim is an automatic time tracker for macOS and Windows: it records the day on your own computer, matches each stretch of work to a task in Jira, Asana, Trello, Linear or GitHub, and writes the timesheet when you confirm it.
I make Planim. The rest of this page follows a working day from raw activity to a finished timesheet, with Planim as the worked example.
What an automatic time tracker records
The tracker watches the computer, and the computer already knows most of what a timesheet needs. This is the full list for Planim:
| Signal | What it tells the tracker |
|---|---|
| Focused app and window title, checked once a second | Which tool, which document, which ticket |
| Browser tab URL and title on macOS (Safari, Chrome, Arc, Edge, Brave) | Which issue, pull request, doc or page |
| No input for 3 minutes, and screen lock | Where work stopped, so lunch is not billed |
| File edits and the git branch | Which code, and the issue key in the branch name |
That is enough to rebuild a working day, gaps included. There are no screenshots, keystrokes, camera frames or locations on the list, and a time tracker without screenshots loses nothing a timesheet needs.
From raw activity to a timesheet: three stages
Every automatic time tracker captures. Most stop before matching.
Capture. The tracker saves an event each time the focused window, tab or file changes.
Episodes. Events are cut into stretches of continuous work. Forty minutes in the editor with two glances at Slack is one episode. Three minutes without input ends it.
Matching. Each episode is assigned to a task. This is the stage that produces a timesheet, and it is the hard one. Clockify's Auto tracker and Toggl Track's Timeline stop before it: they show a list of apps and sites, and you turn it into entries by hand (Planim vs Clockify, Planim vs Toggl Track). Planim does the matching itself. Ten signals on your machine vote on a task for every episode, two of them small language models, and the inspector shows each vote with the text it was based on. The details are in what the AI in an AI time tracker does.
Matched episodes merge into drafts on a day timeline: task, start, end, duration. Those drafts are the timesheet, and they are ready before you open the app.
What changes when the timer goes away
A timer measures the tasks you remembered to start it for. The ten-minute review, the call that ran over, the hour of reading that fixed the bug: none of them had a timer running, and at six in the evening they are gone. People fill the hole from memory, and the timesheet becomes an estimate.
An automatic time tracker starts from the other end. The whole day is recorded, so the question at six is which task each block belongs to, and with matching most blocks arrive already answered. A day with fifteen drafts looks like this in Planim: ten come in with high confidence and take Enter each, four are flagged for a look, one is not work. The confirmed ones take a few seconds each, and the hours add up to the day you had.
Where the record is kept
A day of window titles is a detailed document, so the first question about any tracker is where it stores it. Cloud trackers such as Timely and Rize upload the record to the vendor's servers and analyze it there.
Planim keeps it in a SQLite file on your computer. The matching models run on your CPU. There is no Planim account and no server-side dashboard that anyone could open. Three things go over the network: worklogs you push to your trackers, a one-time download of the two models, and sign-in to a tracker through a relay that deletes the tokens once the app has them. Local AI time tracker walks through each of those.
What you get at the end of the day
You review the drafts with single keys. Enter confirms a proposal, R reassigns it to another task, the I key marks it as not work, and N jumps to the next one that needs attention. Proposals under 80 percent confidence are flagged so they get a second look.
Then you press Push worklog. Confirmed drafts become worklogs on the Jira issues, time entries or comments in Asana, and comments on Trello cards. Linear and GitHub supply tasks to match against. Reports total the day or the week by task, project and category and export to PDF.
Nothing is written to any tracker before that button. Your manager sees worklogs authored by you, exactly as if you had typed them.
Six questions for any automatic time tracker
- Does it stop at a list of apps, or does it assign the time to tasks?
- Does it show why it picked a task?
- What does it do when it is not sure?
- Does the activity record stay on my machine? In which folder?
- Does it write to my task tracker, and only after I confirm?
- Screenshots or keystroke logging, on any plan?
Planim's answers: it assigns every episode to a task, shows every vote behind the choice, flags anything under 80 percent, keeps the record in one local file, writes only confirmed worklogs, and records neither screenshots nor keystrokes. For a tool-by-tool comparison see the best automatic time trackers of 2026.
Planim is in development for macOS 14 and Windows 10/11. Leave an address below and you will hear when the automatic time tracker ships.