The tracker for a professional and their agentic team.

For anyone who tracks time

A timer first.
Agents are optional.

You do not need an agent to use TickTrac. It tracks your hours, shows where they went and gets them ready to bill.

For your agents

Agents file their own hours.

When an agent finishes a task it posts one entry: what it did, for which client, when it started and when it stopped, and how long a person would have needed. Its true time is what bills.

curl -X POST https://ticktrac.com/api/v1/entries \
  -H "Authorization: Bearer $TICKTRAC_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"entries": [{
    "ref": "pr-482-review",
    "name": "Review PR #482: stream export rows",
    "client": "Atelier Nord",
    "categories": ["Development"],
    "tags": ["programming"],
    "started_at": "2026-10-26T09:14:00+01:00",
    "ended_at": "2026-10-26T09:59:00+01:00",
    "source": "agent",
    "human_equivalent_hours": 2.5
  }]}'

{
  "results": [{
    "ref": "pr-482-review",
    "id": 5312,
    "created": true
  }],
  "errors": []
}
# same ref again → "created": false, nothing written

Set up your agent

How to instruct your AI agent to start working with TickTrac.

Copy this prompt into your agent’s instructions. It works with Claude, ChatGPT, Cursor or any agent that can send a web request. After each task, the agent logs its work for you.

What you need first: a TickTrac account, and an API token from Settings → Calendar & API.

Log your finished work to my time tracker, TickTrac.

When you finish a task, send one entry for it.
Send a POST request to https://ticktrac.com/api/v1/entries
with these two headers:

  Authorization: Bearer YOUR-TICKTRAC-TOKEN
  Content-Type: application/json

and this body:

{"entries": [{
  "ref": "a new id for this task, e.g. 2026-10-27-fix-login",
  "name": "what you did, in one short line",
  "client": "CLIENT NAME",
  "categories": ["one category from the list below"],
  "tags": ["one kind of work from the list below"],
  "started_at": "when you started, e.g. 2026-10-27T09:14:00+01:00",
  "ended_at": "when you finished, in the same format",
  "source": "agent",
  "agent": "your own name, e.g. Claude"
}]}

Rules:
- Give every task a new "ref". If a request fails, send it again with the same "ref". TickTrac never logs the same "ref" twice.
- Send the real times you started and stopped, with the timezone. Never an estimate. Never a time in the future.
- Put each entry in one category from this list: Development, Design.
- Tag each entry with one kind of work from this list: docs, programming.
- Never work out a ratio or a human-time estimate. I set those in TickTrac.
- If the answer lists an error, tell me what it says.

Replace the token, the client name and the two lists. Then set a ratio on those tags, or on a category, in Settings → Agents, for example docs 1:10 and programming 1:60. A tag, category or client can take a ratio once it exists: after the agent’s first entry, or once you have used it yourself.

The guards

Made so an agent can’t make a mess of your month.

What you see

One client, one file, both kinds of hours.

Your rows and your agents’ sit in the same export, the agent rows marked. Both bill true time at the client’s rate. Leave the agent work out in one click, and the totals follow.

Questions

Short answers.