How it works
MCP has two sides. An MCP server, published by the owner of a system, describes the tools it offers, such as 'book a desk' or 'list available rooms', and the data it can return. An MCP client, built into an AI assistant, discovers those tools and calls them when a user's request needs them.
Anthropic introduced MCP as an open standard in November 2024, and many AI assistants and business systems now support it. Because the protocol is shared, one server works with any compatible assistant, instead of one integration per assistant.
Security depends on the server. Well-built servers use per-user sign-in, such as OAuth, so the assistant acts with the user's own permissions and never beyond them. Administrators decide whether the server is enabled and who may connect.
For workplace systems, MCP lets people book a desk or a room, invite a visitor or check in from the assistant they already use, without opening another app.
Examples
- An employee asks Claude for a desk near the design team on Tuesday and a room for four at 2 pm; the assistant calls the workplace system's MCP tools and confirms both bookings.
- A host asks Copilot to invite a visitor for tomorrow; the MCP server creates the invitation and the visitor system notifies the host when the visitor arrives.
Common mistakes
- Connecting servers that use a shared service account, so the assistant can do more than the user could.
- Assuming every MCP server offers the same tools; each server defines its own.
- Skipping admin review of which assistants and users may connect.
How Kadence helps
The Kadence MCP server is in early access and included in WorkOps Team, WorkOps Pro and SpaceOps; an admin must enable it. It lets people book desks, rooms, parking and visitors and check in from Claude, ChatGPT or Copilot, with per-user OAuth 2.1 sign-in and the same permissions they already have in Kadence.
Kadence MCP · Integrations · Security
Frequently asked questions
What is an MCP server?
An MCP server is the part of a system that speaks the Model Context Protocol. It lists the tools and data the system offers to AI assistants, such as booking a room, and handles the calls assistants make, checking the user's sign-in and permissions each time.
Who created the Model Context Protocol?
Anthropic created MCP and released it as an open-source standard. It is maintained as an open project, and many AI assistants and software vendors now support it, so one server can work with several compatible assistants.
Is MCP secure?
The protocol carries requests; security comes from how each server is built and configured. Look for per-user sign-in such as OAuth, actions limited to the user's existing permissions, admin control over enabling the server and a record of what assistants did.
What can MCP do for workplace teams?
It lets employees book desks and rooms, invite visitors and check in from the AI assistant they already use, instead of a separate app. For workplace teams, it means fewer tools to roll out and one set of permissions governing every assistant.
Related terms
- Agentic workplace A workplace where AI agents carry out tasks, such as booking space and drafting plans, with people approving decisions.
- AI workplace assistant A chat-based assistant that books desks and rooms and handles workplace tasks from a plain-language request.
- Workplace API A programming interface that lets other systems read and write workplace data such as spaces, bookings and check-ins.
- Microsoft Teams and Slack booking Booking desks and rooms, seeing who is in and checking in from inside Microsoft Teams or Slack.
Sources
- Introducing the Model Context Protocol, Anthropic
Kadence editorial team · Last reviewed