The vibe-code agent on your server can do more than answer questions about Kimai — it can run scripts on a schedule, call Kimai’s JSON API, and wire your tracker to the other tools your team lives in. The ten workflows below are multi-step automations: each one reads data, makes a decision, and acts. Start with the reminder loop, then take the ones that save you an hour a week.

1. The 4 pm missing-hours nudge

Daily logging dies the moment people forget a tracker exists. This workflow runs on a weekday cron, asks the Kimai API for each active user’s logged hours today, and sends one short digest — name plus gap — covering only the people who are short. It is deliberately nagging-light: a single message, and only to those who need it. It can also skip public holidays by reading the working calendar, so nobody gets pestered on a day off. Say: “Every weekday at 4 pm, check who has logged fewer than four hours today and post a reminder listing their gaps.”

2. Friday project digest to your team chat

Weekly totals are the cheapest management information you own, but exporting them by hand every Friday is a chore nobody keeps up. The agent writes a script that pulls this week’s hours per project and per user from the API and posts a formatted summary to your Slack, Discord, or Matrix webhook. It can include the change against last week and flag projects with zero time. Say: “Every Friday at 5 pm, post hours per project and per user for this week to our Discord channel, with the change from last week.”

3. Monday draft-invoice run

Kimai ships a console command for batched invoicing that is built to run from cron. A scheduled agent script regenerates draft invoices for every customer with unbilled billable entries from the previous month, writes the PDFs to a folder, and emails you the list with totals. You review and send — nothing goes to a client without a human looking at it. Say: “Every Monday, create draft invoices for any customer with unbilled billable entries from last month, save the PDFs, and email me the list with totals.”

4. Timesheet close-out and Monday review

Pair the lockdown feature with a weekly review and invoicing gets boring in the best way. Every Friday evening the workflow locks the just-finished week after its grace period; Monday morning it exports a CSV of everything logged and sends it to you for approval before any invoice exists. Two small scripts, one reliable routine. Say: “Close last week’s timesheets every Friday at 6 pm, and Monday morning send me a CSV of everything logged last week so I can approve it.”

5. Turn calendar meetings into entries

Meetings are the time everyone forgets to log. A morning job reads yesterday’s events from a shared work calendar — over an ICS feed or the calendar API — and creates candidate entries mapped to a customer or your Internal project by matching organizer and title. You confirm or delete them; over time the mapping rules learn your recurring patterns. Say: “Each morning, pull yesterday’s events from the team calendar and turn anything over 15 minutes into entries on Internal for me to confirm.”

6. Start and stop the timer from chat

Once the API is reachable, the chat itself becomes a time entry surface. The agent sets up a small wrapper: saying “start” with a project and activity name creates a running timesheet, and “stop” closes the open one and reports its duration. This is the lowest-friction capture you can give people who already live in chat. Say: “Set it up so ‘start Acme Website Development’ begins a running entry and ‘stop’ ends it and tells me how long it ran.”

7. Payroll cross-check at month end

Logged hours and contracted hours drift, and the drift costs money in both directions. A monthly script compares each person’s total logged time against their contract, flags anyone more than ten percent over or under, and attaches the per-user breakdown so the follow-up conversation starts with facts instead of vibes. Run it a week early as a soft warning and the month-end surprise disappears. Say: “On the last day of the month, compute each person’s logged hours against their contracted hours and email me anyone who is 10 percent over or under, with the details.”

8. Project budget guardrails

Fixed-price projects live or die on burn awareness. A nightly job totals hours per project against a budget you keep in a config file or a custom field, and emails you when a project crosses 80 percent and again at 100 percent — with days remaining if you gave it an end date. Say: “Check logged hours on the Website project against its 400-hour budget every night, and email me when it crosses 80 and 100 percent.”

9. Backups with a real restore test

A backup you have never restored is a hope, not a backup. The agent sets up a nightly job that dumps the Kimai database, archives the files, prunes anything older than two weeks, and — the part everyone skips — restores the newest dump into a scratch database once a week to prove it parses. Say: “Back up the Kimai database and files nightly at 2 am, keep 14 days, and once a week restore the latest dump to a scratch database and confirm it worked.”

10. Nightly auto-tagging rules

Tags only stay useful if they are applied consistently, and humans are not consistent. A small rules file — description contains “support”, activity is Meeting, project is Internal — runs nightly through the API and tags matching entries, so your billable and project reports stay clean without anyone remembering to do it. The rules live in a plain file, so changing them later is a one-line edit, not a rewrite. Say: “Every night, tag entries that mention support in the description as support-call, and tag anything over eight hours for review.”

Automations are promises your server keeps while you do other work, so start with one and read its output for a week before adding the next. Kimai on OpenSysLab