DELog inSign up
Track a day with Claude Code and Teetrack

Track a day with Claude Code and Teetrack

By Tristan Doehl · · Updated

Claude can use Teetrack tools while you work in the same conversation. A useful first workflow is a single day of tracking: connect the account, start and stop a timer, then compare the resulting entries with the work you completed.

Connect Claude Code to the hosted server

For Claude Code, register the HTTP endpoint with this command. It connects to the hosted Teetrack service rather than starting a local copy of the application.

claude mcp add --transport http teetrack https://teetrack.it/api/mcp

Open Claude Code and use /mcp to inspect the connection and complete authentication when prompted. Anthropic’s MCP documentation describes the configuration scopes (local, project, and user) and the authentication options. For Claude’s web and desktop apps, follow the separate Claude setup guide.

After connecting, ask Claude to list your Teetrack projects. Check a familiar name against the web application. This small read request is useful because a successful login alone does not establish that you selected the account or permissions you intended.

Start with a specific project and task

An illustrative morning request is: “Start a timer for checkout testing on the Acme redesign project.” Claude can look up the project and start the timer through MCP. Inspect the result for the project and description; if the name is ambiguous, identify the correct project before continuing.

At the next task change, ask it to stop that timer. Without a timer ID, stop_timer stops the most recently started timer. If several timers are running, ask for the list and identify the intended timer before stopping it.

Correct a missed session with explicit times

If you forgot to track a call, supply the date and the actual start and end times. For example: “Log the Acme planning call on 27 February 2026 from 14:00 to 14:45, Europe/Berlin, on Acme redesign.” This gives the assistant enough information to construct a completed entry without guessing its length.

Review the entry afterwards. A wrong duration or description can be edited, but deletion is not a general undo mechanism. Ask to inspect the matching entries before changing several records, and use the assistant application’s approval settings for operations that remove data.

Review the day before using its totals

At the end of the day, ask for the completed entries grouped by project, including their durations and billable status. Asking for completed entries matters when calendar sync is on, because an unfiltered list also contains unconfirmed calendar drafts, which the response marks with draft: true. Compare the list with your work and stop any timer that should have ended. If an operation failed, resolve it before requesting a billing summary.

The assistant’s explanation and the stored records serve different roles. The records are what Teetrack uses in reporting, so check them when a conversational summary seems inconsistent. A fresh query is more reliable than asking the assistant to reconstruct earlier results from the conversation.

Keep the workflow small enough to inspect

The MCP introduction explains the account and data boundaries behind this connection. Once a single day is being recorded correctly, you can extend the same routine to a weekly review. Clear targets and checked results remain useful even when the request fits into one sentence.

Tristan Doehl

Tristan Doehl

Tristan is the founder of Teetrack and a software engineer based in Aachen, Germany. He started Teetrack as his own tool for tracking billable hours as a developer and has been building it into a full time-tracking product for freelancers and small teams ever since. He writes about time tracking, freelance billing, and AI-assisted workflows.