How it works
The Model Context Protocol (MCP) is described on its own site as an open-source standard for connecting AI applications to external systems. An AI application can use it to reach data sources such as files and databases, tools such as search or calculators, and prepared workflows. The site compares it to a USB-C port: one standard way to plug an assistant into many different things.
Two roles take part:
- The client is the AI application (a chat app or a coding assistant). It connects to servers and lists what they offer.
- The server is your program. It declares tools, each with a name, a description and an input schema, and runs them when asked.
The assistant reads the tool descriptions, decides when one fits the request, fills in the inputs and shows you the result. Nothing is hard-coded on the assistant’s side.
Example: a support request
An illustrative exchange, not a transcript. A seller asks an assistant that is connected to a license-management server: “Find Ada’s license and extend it by a month, her subscription renewal was late.”
- The assistant calls a lookup tool with her email and finds one license that expired two days ago.
- It reports what it found and the new expiry date it proposes, and asks before changing anything.
- After you agree, it calls the extend tool and reports the license’s new expiry.
The same steps through a REST API would need a script: find the endpoint, handle the pagination, send the right body. The assistant can do it because the server told it what exists.
MCP server versus a REST API
| REST API | MCP server | |
|---|---|---|
| Caller | Your code | An AI assistant, through a client |
| Discovery | You read the docs | The server lists its tools |
| Best for | Fixed integrations and backends | Ad hoc work: lookups, support, one-off fixes |
| Review before action | Your code decides | The client can ask you to approve each call |
An MCP server usually wraps the same operations as an existing API, so it rarely adds new capabilities. It adds a way to use them in conversation.
How Clavem handles it
Clavem runs an MCP server at https://clavem.io/mcp. It lets an assistant manage your licenses: look licenses up, issue one, extend one, and revoke one. It authenticates with a management API key from your dashboard, the same kind of key the REST management API takes. The key is shown once and can be revoked at any time, so give the assistant a key made for that purpose and delete it when you stop using it. Setup steps for specific clients are in the MCP docs, and the REST equivalents are in the API reference.
One piece of advice: a revoke tool in an assistant’s hands can end a paying customer’s access. Configure your client to ask for approval before any tool that changes data, and read the proposed action before you accept it. Looking things up is low risk; issuing, extending and revoking are not. Why you might want this at all is covered in manage licenses with AI and MCP, and the dashboard equivalent in license manager.
Questions
Is an MCP server the same as a REST API?
No. A REST API is for programs that already know its endpoints. An MCP server describes its tools to an assistant, which can then choose to call them during a conversation. Clavem's server works with the same kind of management API key as its REST API.
Can an assistant revoke licenses through Clavem's MCP server?
Clavem's server offers tools to look up, issue, extend and revoke licenses. Whether an assistant may call them without asking depends on how you configure your client, so require confirmation for destructive actions.