MCP server: the shift optimizer inside your AI assistant

Why use it

A language model can write a roster that looks right and still gives someone six hours between a night shift and a day shift. With this server the assistant turns what you tell it into the input, and the roster itself is computed by the same optimization solver (mixed-integer programming, HiGHS) that runs in the web version: minimum rest, consecutive nights, skills and availability always hold, and any post that cannot be covered is named instead of silently dropped.

How to connect

Claude: at https://claude.ai/customize/connectors add a custom connector with the URL above. No OAuth needed.

Claude Code: claude mcp add --transport http shift-scheduler https://staff-scheduling.kindoflost.com/mcp

Other assistants and clients: anything that accepts a remote MCP server over HTTP; use the same URL.

What to ask

For example: “Build an October roster for my ward: 9 nurses, 7am–7pm and 7pm–7am shifts, two on days and one on nights, Ana can't do weekends, nobody more than three nights in a row.”

Tools

• list_shift_examples — the five built-in problems (hospital ward, 3-shift plant, security post, fire station, 24/7 service desk).
• get_shift_example — the full input of one example, to edit into your own.
• build_shift_roster — solves the roster and returns who works which shift each day, coverage, unfilled posts, hours per person and warnings.
• get_shift_roster_result — collects a long solve that did not finish inside the first call.

Limits

Up to 60 people, 42 days, 12 shifts and 12 posts. A run takes 5 to 60 seconds; the server solves one roster at a time, so at a busy moment the assistant may be asked to retry in a minute.

Privacy

What the assistant sends is used only to compute the roster: it is held in memory while the solve runs and for at most 15 minutes after, then discarded. Each run writes one anonymous log row, with no names and no roster data. Details in the privacy policy.

Support: mario@kindoflost.com.