TimeCampus
AI & Agents2 illustrations

In-Context Assistant and Agent Access

Ask about the screen you are on, or let external AI agents act on your time data, always under your own permissions.

These images are illustrations of the concept, not screenshots of the actual product.

Overview

This concept brings AI into TimeCampus in two complementary ways. Inside the app, an assistant panel understands which screen the person is on and can answer questions or take actions there, such as starting a timer for the project being viewed. Outside the app, external AI agents can reach TimeCampus through one governed gateway that exposes the product's capabilities as grouped tools.

The risk with AI in a time product is trust. An assistant that sees everything, or an agent that can act with broader rights than its user, would undo the privacy-first stance the product is built on. The design addresses this directly: the assistant shares only minimal context about the current screen, and both the assistant and external agents act with the person's own identity and permissions. Destructive actions requested by an agent pause for explicit confirmation.

The first illustration shows a project detail page on its Time tab, with logged, budget, billable and status figures, and the TimeCampus Assistant open alongside. A context chip names the area, project and tab. The person asks the assistant to start a timer for the project, and the reply confirms the timer started and that it used the person's own permissions, with nothing bypassing them. Suggested questions offer help with the current screen, this week's time entries and how the focus score works, and a footer states that only the route, entity, tab and project are shared. The second illustration shows the Agent access page: a connected status, counts of tools available, calls today and calls blocked for confirmation, and a Tool groups table covering areas such as time entries, meetings, location, Pomodoro, resourcing, configuration, approvals, expenses, focus and projects, with tool counts, destructive-tool counts and last use. A How access works card summarizes the rules, and a confirmation dialog holds an agent's request to delete time entries until the person confirms or cancels.

Because both surfaces are designed to act through the same permissions as the person, they sit alongside the product's privacy and consent controls, and they reach across timers, projects, approvals and other concepts rather than replacing them.

What this concept shows

  • An assistant panel that opens beside the current screen and shows the context it is using
  • Natural-language actions such as starting a timer for the project being viewed
  • Replies that confirm an action ran with the person's own permissions
  • Suggested questions tailored to the current screen
  • A footer stating that only the route, entity, tab and project are shared
  • An Agent access page with connection status and counts of available tools, calls today and blocked calls
  • A Tool groups table with tool counts, destructive-tool counts and last-used times
  • A confirmation dialog that holds destructive agent requests until the person confirms

How it works

  1. Open any screen, such as a project's Time tab, and open the TimeCampus Assistant beside it.
  2. Check the context chip to see which area, project and tab the assistant is working with.
  3. Ask a question or request an action, or choose one of the suggested questions.
  4. Review the reply, which confirms what ran and that it used your own permissions.
  5. Open Agent access to see the connection status, available tool groups and today's calls from external agents.
  6. When an agent requests a destructive tool, review the confirmation dialog and choose Confirm and run or Cancel.

Who it's for

  • Individuals who prefer asking over navigating
  • Team members who log time across many projects
  • Developers and power users connecting AI agents to their time data
  • Workspace administrators who need oversight of agent actions

Illustrations

2 illustrations of this concept. Select one to view it full size.

Assistant Beside a Project's Time Tab

The assistant knows which screen you are on, acts there, and says exactly what it shares.

This illustration shows a project detail page for a sample project marked Active, with Overview, Tasks, Time and Budget tabs and the Time tab selected. Summary cards show logged hours, budget, billable percentage and status, above an overview table with progress bars comparing time against budget. The TimeCampus Assistant panel is open on the right with Assistant, API Calls and General tabs. A context chip identifies the screen as the project's Time tab. The person asks the assistant to start a timer for this project, and the reply confirms the timer started and states that it used the person's own permissions, with nothing bypassing them. Suggested questions follow: what can I do on this screen, show my time entries for this week and how Focus Score works. An input invites questions about the screen, and a footer notes that only the route, entity, tab and project are shared. The design demonstrates contextual, permission-bound help.

Governed Agent Access and Tool Groups

External agents run as you through one gateway, and destructive tools wait for your confirmation.

This illustration shows the Agent access page, subtitled as the one governed gateway through which external agents reach TimeCampus, with a Connected badge. Three cards count tools available, calls today and calls blocked for confirmation, using sample figures. A Tool groups table lists groups such as time entries, meetings, location, Pomodoro, resourcing, configuration, approvals, expenses, focus and projects, with columns for the number of tools, the number of destructive tools and when each group was last used. A How access works card explains that agents run as you with your identity and permissions, that every call goes through the public gateway and that destructive tools need explicit confirmation. In the foreground, a Confirmation required dialog reports that the agent requested a destructive tool to delete three time entries and that nothing has run yet, and it offers Confirm and run or Cancel. The design demonstrates human-in-the-loop control over AI agents.

Topics

  • AI assistant for time tracking
  • context-aware in-app assistant
  • start a timer with natural language
  • AI agent access to time data
  • MCP tools for time tracking
  • governed agent gateway
  • human confirmation for AI actions
  • AI agent permissions
  • privacy-first AI assistant

Related concepts

All TimeCampus concepts
Get startedBrowse all concepts