AI that reads your day, not your screen
TimeCampus records intervals, never screenshots. Naming the project and the life area behind each hour is the product itself — and nothing reaches your timesheet unless you say so.
Manual timers fail because you forget them and monitoring tools produce data nobody trusts, so TimeCampus captures plain intervals instead — an app, a title, a stretch of time, no pixels.
That stream is worthless until something names the project and the part of your life behind each hour, so plain rules run first, a model only refines, and you decide what counts.
Everything below is labelled with where it actually stands in TimeCampus today — shipped, in progress, or on the roadmap.
AI-native by architecture, not bolted on
Four architectural choices that make AI a first-class part of TimeCampus rather than a chat window on the side.
A log of hours is not an answer
TimeCampus stores an app, a title and a span of time. Naming the project and the life area behind each one is the only output worth reading.
Every label says who decided it
Each labelled hour shows whether a rule, a model or you chose it, which model it was, and how sure it was. You can always tell the three apart.
Scored on balance, not on output
The alignment score compares the shape of your week to the one you asked for, so working longer cannot raise it. A calm week can beat a frantic one.
Switch a source off and nothing is kept
Withdraw a source and the batch is refused before anything is written, and you are told how much was refused. Your personal analysis stays yours.
From intent to a governed action
Walk the path a request takes. Select any stage to see what happens there and what backs it.
Stage 1 of 5 · Capture
Consent is checked before anything is stored
Your tools send batches of activity — an app, a title, a link, a span of time. No screenshots, no keystrokes.
If you have switched a source off, the batch is refused outright and the refusal is counted, so nothing further along can ever see it. Repeats of the same activity are ignored, so a tool re-sending yesterday cannot inflate your week.
What the AI in TimeCampus actually does
Filter by delivery status, then open any capability for the detail and what backs it. No capability is listed as shipped without something in the product behind it.
Where the AI shows up
The places AI meets the work in TimeCampus — and how far each one has actually got.
In-app assistant
ShippedA floating assistant on every signed-in screen, told only where you are and nothing about who you are or what you use.
- Present everywhere you work, not parked on one separate AI page
- Knows the screen and the item you opened — not your device or your browsing
- Cannot do anything you could not do yourself, and keeps no record of its own yet
See it as a real scenario
Every AI capability above shows up in a concrete TimeCampus story. Open one to read the full walk-through.
Let TimeCampus name the project behind each hourYour own rules first, a model second, and neither one allowed to touch your timesheet without you.Read the story
Weekly suggestions that will never tell you to work moreFour plain rules read your own week, and each card shows the numbers it came from.Read the story
Insights that show their evidence and admit what they missedEvery score reports how much of the week it could account for, and returns nothing rather than a made-up number.Read the story
Find the hours that quietly leaked out of your weekWaiting, commuting, unlabelled activity and untracked gaps — with sleep and chosen time never counted as waste.Read the story
Ask the assistant about the screen you are already onIt knows the section, the item and the tab in front of you — and has exactly the permissions you have.Read the story
Decide what the intelligence is allowed to seePer-source consent honoured before anything is stored, and personal context an employer simply cannot read.Read the story
Give your coding agent a governed door into your time163 MCP tools, your permissions unchanged, and a confirmation stop before anything destructive runs.Read the storyAI you can actually let near your data
An AI-native product has to be governable. Here is where TimeCampus stands on each control — including the parts still being built.
Permissions checked before the AI, not by it
Personal analysis and recommendations are yours alone. An employer or workspace account cannot reach them, and cannot even confirm that yours exist.
Sources you switch off are refused
Nothing from a withdrawn source is stored, and you are told how much was refused. Honest limit: this stops new capture, and does not yet remove what was already kept.
No guess ever lands on an invoice
A suggestion sits beside the hour, never on it. Only your acceptance or your own edit writes the project that drives billing — and your edit is marked as yours for good.
Destructive agent actions stop and ask
Tools marked destructive will not run without an explicit confirmation. The open work is checking that marking tool by tool, with tests that hold it in place.
A record of every model call
There is deliberately none today, since a suggestion you approve is not an autonomous act. A shared record is planned, and comes before anything may run unattended.
Quality tests for labelling and the assistant
Neither has a fixed set of expected answers yet, so a drop in quality would be noticed by you rather than caught by us. Building both is planned.
Models
TimeCampus uses one provider — the first one configured — with a single attempt and a twelve-second timeout, and labels every result with the model behind it. There is no multi-model routing and no bring-your-own-key yet; with none set, the rules still work.
- Google Gemini
- OpenAI
- Anthropic
The honest answers
Only on one path, and only if a provider is configured. When you ask for a project suggestion, the app, title and link for that one hour, plus your own project names, can be sent to a single model so it picks one of your names. There are no screenshots or keystrokes to send. With no provider set, that step is skipped.
Put TimeCampus’s AI to work
TimeCampus records intervals, never screenshots. Naming the project and the life area behind each hour is the product itself — and nothing reaches your timesheet unless you say so.