Connecting an AI Application to Time Doctor (MCP) - Beta
Note: The AI connection (MCP) is available on all Time Doctor Subscription plans and can currently be accessed by Owners, Admins and managers.
TL;DR:
The AI connection (MCP) links an AI assistant — such as Claude or ChatGPT — to Time Doctor data for plain-language reporting. Ask about hours tracked, productivity, apps and websites, idle time, meetings, projects, attendance, leave, and teams. Access is read-only and limited to the assigned user role.
What Is the AI Connection (MCP)?
Connect an AI assistant — such as Claude or ChatGPT — to Time Doctor data using the AI connection (MCP), and ask questions in plain language. Skip opening reports: ask a question directly, and the AI assistant answers using live Time Doctor data.
AI connection (MCP) access is read-only. The connection can read and report data but cannot change, add, or delete anything in Time Doctor. Data shown is limited to what the assigned user role is permitted to see.
What to Ask
Use the categories below as a starting point. Follow up on any answer to narrow it down by person, team, or date range, or to break a total into more detail.
| Topic | Try Asking |
| Hours tracked | "How many hours did the Sales team log this month?" |
| Productivity | "What's the Support team's productivity score this week, and what's driving unproductive time?" |
| Apps & websites | "Which apps and websites are people spending the most time on? Are they using AI tools?" |
| Idle & activity | "What's the idle time for my team this week, and who had the most?" |
| Meetings | "How much time did the team spend in meetings, and what percentage of work was meetings?" |
| Projects & tasks | "Which projects took the most time this month? Break it down by task for Ana." |
| Attendance | "Who didn't track any time today? Who's late?" |
| Leave & time off | "Who was on leave this month, and how much has Carlos taken?" |
| People & teams | "List the people on the Sales team, and who this manager manages." |
| Trends | "How do hours tracked trend across the week? Show productivity over the last month." |
It can also answer harder questions that are tricky to pull from standard reports — for example: "We have no-meeting Fridays. For the Marketing team in June, who's actually keeping Fridays meeting-free? Give me a visual report." Or correlate Time Doctor activity with data you provide (like closed deals) to see what work patterns track with results.
Note on attendance: "Who's late?" works when Settings / Schedules is set up; otherwise include a start time in your question.
These are just examples, not limits. The AI connection can work with any Time Doctor data your role can see, so feel free to experiment and ask questions your own way — combine topics, phrase it however's natural, or try use cases beyond this list.
Whose Data Is Visible
The connection uses the assigned Time Doctor access level, so visibility follows existing permissions. Owners and Admins can ask about everyone in the company.
Tips for the Best Answers
- Name a date or range — “today,” “last week,” “June,” or “between April 1 and 30.”
- Name the person or team being asked about.
- Mention a timezone for day boundaries in local time — for example, “in Eastern time.” by default the answers are returned according to your local timezone set in the Time Doctor account.
- If the AI assistant doesn't automatically use Time Doctor, name it directly — for example, “Using Time Doctor, show hours tracked this week.”
- Ask follow-up questions to drill in — start broad (“How is the team spending time?”), then narrow (“Break that down by task for Ana.”).
Understanding Data Limits
To keep responses fast and reliable, the AI connection (MCP) works within a few limits. Knowing these limits helps with asking questions that return complete and accurate answers.
- Group size: A question about a group or team covers up to 200 members at a time.
- Date range: Ask for up to about one year in a single request. Time Doctor keeps roughly two years of history.
- Amount of data: Results come back in batches. A row is one line of data, similar to a line in a spreadsheet, and each row represents one record. Detailed data, such as Web & App Usage, creates many rows because each website and app used by a person is stored as a separate record. One person can generate hundreds of rows in a single day.
For larger analyses, work in batches. Avoid asking for hundreds of users across a long period in a single request. Even when the request is technically within the limits, that amount of data can cause a timeout or hit a data limit, which may return an incomplete answer.
Instead, break the request into smaller parts and build the result step by step. For example, to analyze 500 users over a full month, do not request everything at once. Start with a smaller batch, such as 50 users for one week, let the assistant return that batch, then continue with the next group and the next week. After gathering all records, ask the assistant to combine them into one analysis. This takes a few more steps, but it returns more complete and accurate results than a partial response.
If a limit is reached, narrow the request by using a shorter date range or fewer users. Another option is to ask for a summary or top results instead of every row, such as top 10 apps. If needed, continue in smaller batches until all required data is gathered.
Should there be any inconsistencies or concerns regarding the article, contact support@timedoctor.com for prompt assistance.