Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external systems through a uniform JSON-RPC interface. The Dime.Scheduler MCP server exposes the most common scheduling operations - listing tasks, finding available resources, creating and rescheduling appointments - as MCP tools that any compliant client can call.

What you can do with it

With an MCP-capable assistant connected to Dime.Scheduler, you can:

• Ask in plain language - "Who is available next Tuesday afternoon for a 2-hour electrical job in Antwerp?"
• Schedule and reschedule appointments - create, move, reassign, or cancel work orders directly from chat.
• Inspect the planning - pull a resource's calendar for the week, search appointments, or look up open tasks.
• Discover the right resource - match jobs to skills, location, and availability through the recommendation tool.
• Drive automations - embed Dime.Scheduler tools in larger AI workflows alongside calendars, CRMs, and ticketing systems.

How it fits in the integration landscape

Connector	Audience	Style
Business Central / NAV	Back-office sync	Bidirectional, transactional
Exchange	Calendar sync	Bidirectional, scheduled
Webhooks	Custom downstream systems	Outbound, event-driven
Zapier / Logic Apps	No-code automation	Bidirectional, workflow
MCP	AI assistants and agents	On-demand, conversational

MCP complements the existing connectors: it does not replace your back-office sync, but it unlocks a new conversational entry point on top of the same data.

Architecture

AI assistant (Claude, Cursor, …)  ──►  MCP client
                                              │  JSON-RPC over Streamable HTTP
                                              ▼
                            https://<env>.api.dimescheduler.com/mcp
                                              │
                                              ▼
                                  Dime.Scheduler API

Every request is authenticated with the same API keys you already use for the public API. The MCP server is stateless - it translates tool calls into Dime.Scheduler API calls and streams the results back to the assistant.

What's exposed

The MCP server exposes three kinds of capabilities. Most of the work happens through tools, but resources and prompts complement them in important ways.

• Tools - actions the agent can call (search, create, reschedule, optimize, ...).
• Resources - read-only data sources the agent can pull as standing context without consuming a tool call.
• Prompts - pre-built workflows the agent can run end-to-end.

Tools

The server exposes tools across the following areas. Each is documented on its own page:

• Tasks & Jobs - list open tasks, search tasks and jobs.
• Resources - search resources, inspect skills and capacity, find available slots, get recommendations.
• Appointments - search, inspect, create, update, reschedule, and delete appointments.
• Optimization - generate optimal route and assignment plans.
• Messaging - send real-time messages and inspect notifications.
• Reference data & utilities - categories, time markers, relative-time parsing.

Resources

Resources are addressable via dimescheduler://... URIs. Reading a resource returns up-to-date data without consuming a tool call, which is useful for keeping standing context (e.g. the current notification stream, the resource roster, the list of categories) cheap.

URI	Purpose
dimescheduler://today/open-tasks	The current set of unplanned (open) tasks. Use this instead of a dedicated tool call for the queue snapshot.
dimescheduler://resources	The complete resource roster with basic info. Use this when you want the full list - for filtered lookups, prefer search_resources.
dimescheduler://resource/{resourceDisplayName}/location	Most recent GPS ping for a specific resource. Templated: substitute {resourceDisplayName} with the resource's DisplayName from the dimescheduler://resources resource. Returns null when no GPS exists.
dimescheduler://recommendation/scorers	The scoring strategies available to get_recommendations.
dimescheduler://categories	All appointment categories.
dimescheduler://time-markers	All time markers.
dimescheduler://notifications/recent	The most recent 50 notifications for the tenant - newest first, including severity, code, message, related entity ids and timestamp. Use as standing context to surface schedule changes, errors, and import warnings.

Prompts

Prompts are pre-built workflows. Each one bundles a sequence of tool calls and resource reads into a guided procedure that the agent walks through end-to-end. They all share one rule: propose, don't act - the agent stops to confirm with the dispatcher before making destructive changes.

Prompt	Purpose
optimize-and-explain	Run field-service optimization for one or more resources over a date range and explain the trade-offs (BEFORE vs AFTER, travel time delta, distance delta, reassignments). Wraps optimize_field_service.
cancel-and-notify	Cancel one resource's appointments for a given date and notify them. Composes search_appointments + bulk_delete_appointments + send_message, with explicit confirmation before any deletion.
find-conflicts	Inspect a date's schedule for double-bookings, impossible travel times, and unmarked travel gaps. Reads dimescheduler://resources and calls get_resource_planning per resource.
audit-resource-day	Inspect one resource's full day - planned appointments, gaps, conflicts, and GPS movement. Useful for end-of-day review or "what did Sarah actually do today?" questions.
balance-workload	Assess workload distribution across resources for a date range and propose rebalancing for outliers. Computes utilization, flags > 90% (overbooked) and < 40% (underbooked), suggests reassignments without applying them.
onboard-task	Guided workflow to schedule a single task: confirm details, find the best resource through get_recommendations, present top candidates, and call create_appointment once the dispatcher picks one.
triage-open-tasks	Walk through the open tasks queue and propose assignments. Reads dimescheduler://today/open-tasks and dimescheduler://resources, then proposes a create_appointment per task without creating anything until the dispatcher confirms.

Next steps

1. Check the Prerequisites.
2. Pick the right endpoint and authentication for your environment.
3. Connect your favorite MCP client - Claude Desktop, Cursor, the MCP Inspector, or your own.
4. Try the example prompts to get a feel for what's possible.