Ask your team's week from Claude or ChatGPT.

Most questions about a team are one line long: who is online, how many hours did we bill this week, was anyone still on that site yesterday. Answering them normally means opening four reports and setting the same date range four times. TimeNova connects to the assistant you already use, so you ask the question instead.

What you can ask

Thirteen read-only questions are exposed, and between them they cover every report in the product. In practice you do not learn their names; you ask in English and the assistant picks.

• Hours and money. Tracked, active and idle time per person for any range, with breaks, manual time and cost where your account uses them. The same figures as the Hours worked report.
• The shape of a day. One person's timeline told as a story, plus meeting time, media time and the low-input stretches that explain a quiet afternoon.
• Apps and sites. What was used and for how long, and when somebody was on a particular site, visit by visit, with the pages they opened inside it.
• Attendance. Who started late, who is absent against their own schedule, day by day.
• Activity flags. Anything that looked unusual, described rather than accused, with the captures from that exact window if you want to look before you speak to anybody.
• Right now. Who is online, who is away, who is on a break, and what they have in front of them.
• People and projects. Who is on the roster and in which group, and where the hours went by project.

What it deliberately cannot do

This is a monitoring product, so the interesting part of an AI connector is not what it can reach. It is what it cannot.

• It cannot change anything. Read only is structural, not a promise: the connector may only issue GET requests, and only against an allowlist of report endpoints. It cannot change a setting, start or stop a tracker, invite anybody, or reach the company key.
• It cannot see more than you can. The key belongs to one person and carries that person's live permissions. A manager gets their group. Demote somebody and their key narrows the same second.
• It cannot be quiet about it. Every question it asks is written to your activity log with the path and the filters, never the answers.
• It cannot outlive your decision. Revoke the key in Settings and it stops working immediately. Archiving somebody revokes theirs; deleting somebody destroys theirs along with their data.

Which assistants

Claude and ChatGPT are both confirmed against their real runtimes. TimeNova speaks the Model Context Protocol, so any assistant that can add a custom MCP connector can use the same address. Google's consumer Gemini app cannot add one today, so we do not claim it; the same address does work anywhere custom MCP is supported, including Gemini CLI.

Which plan

AI assistants are part of the Max plan. The 14-day trial includes them, so you can try the questions on your own team first.

Why we built it this way

A tracker that can be asked questions is only safe if the asking inherits the same limits as the looking. That is why the connector is not a new API with new permissions: it reaches exactly what a signed-in person reaches, through the same endpoints, with the same role scope and the same transparency rules described in security. Nothing was widened to make this feature possible.