Most directory servers stop at discovery. This one carries the sequence through to a created booking, which changes what a mistake costs: a wrong search returns the wrong plumber, a wrong booking puts an appointment in someone's calendar. The ordering tools are there for that reason — get_refinement_options is documented as something to call before searching, and ask_business_agent requires the exact slug copied from a search result rather than one you assembled. Worth knowing that the npm package is a shim over the hosted endpoint, so the tools come from Qasper's server at call time and the package version says nothing about which tools you will get; a remote-capable client can skip it and attach to the URL directly.
A client for Qasper's directory of local service businesses — barbers, dentists, plumbers, salons, mechanics — with ten tools spanning discovery, pricing, availability and booking. The npm package is a thin stdio shim that proxies to the hosted server at https://qasper.ai/mcp, and pulls its tool definitions live from upstream.
- search_businesses is the entry point, with get_refinement_options listing the dimensions a subcategory supports — specialisations, practice areas, service types and modes — so a search is refined on terms the directory recognises.
- get_business_info and get_services return the profile and the catalogue: service names, durations, price ranges, participant limits and the booking start policy.
- get_pricing applies the business's own rules, including emergency and weekend rates, rather than quoting a list price.
- check_availability returns the windows the calendar is free and the booking policy permits, find_next_available_appointments searches across matching businesses for the soonest slot, and book_appointment creates the record and returns a reference number with a status.
- send_inquiry covers the case that does not fit a bookable service — a custom quote or a described problem — and ask_business_agent routes a question to that business's own agent.
Nothing to supply. It can be attached as a remote HTTP server at https://qasper.ai/mcp with no install, or run through npx for clients that only speak stdio. The upstream endpoint the shim proxies to is configurable.
One command — npx -y @qasperai/mcp-server