A working MCP server you can connect to right now

MCP (Model Context Protocol) is the open standard that lets AI assistants use outside tools and data. This server runs at https://mitc.dk/mcp. Svend Madsen built it, and he builds custom MCPs for businesses the same way.

What a custom MCP and MITC give you

Your AI assistant works with your systems. An MCP server gives an assistant tools it can call: look up a record, list open items, fetch a document. It answers from your data instead of from general knowledge.

One server works in Claude, ChatGPT and Cursor. MCP is an open standard, so the same server connects to any client that supports remote MCP servers. You don't build a separate integration for each assistant.

Scoped and read-only by default. You choose which systems and which actions the assistant can reach. A server starts read-only, and write actions are added only where you decide they belong.

Data and development teams that work with AI. Svend AI-enables data teams and development teams with Cursor, so AI assistants become part of how they build, query and ship every day.

Try it now

Server address: https://mitc.dk/mcp

This server is open and read-only. It needs no login and no API key.

Claude. In Claude's settings, find the connectors section and add a custom connector. Paste the server address and save. Start a new chat and ask: "What can the MITC server do?"

ChatGPT. In ChatGPT's settings, find where you add apps or connectors and add a custom MCP server by URL. Paste the server address. Custom MCP servers depend on your ChatGPT plan and settings. Start a new chat and ask the same question.

Cursor. Add the server to your MCP configuration file (~/.cursor/mcp.json for all projects, or .cursor/mcp.json in one project):

{
  "mcpServers": {
    "mitc": {
      "url": "https://mitc.dk/mcp"
    }
  }
}

Then check that the server is enabled in Cursor's MCP settings, and ask the agent: "What can the MITC server do?"

What you just saw

Every example in mcp_patterns_demo uses sample data. None of it comes from a real company.

Card. The tool returns one item with a few labelled fields, the way a customer, order or ticket lookup would. The assistant gets structured fields, so it can quote them exactly.

List or table. The tool returns several rows with the same columns. A real server would use this to list open invoices, recent orders or failed pipeline runs.

Form-style input. The tool takes named, typed inputs, such as a date range or a status filter. The assistant fills them in from your question, and the server checks them before it runs anything.

Resource. A resource is a document or data item the server exposes for the assistant to read, such as a policy, a schema or a product sheet. The assistant reads it as context. It does not change anything.

Prompt. A prompt is a ready-made instruction the server offers, such as "summarise this week's tickets".

Contact

MITC is Svend Madsen's consultancy in Denmark. If any of these is relevant to you, get in touch:

  • Consulting on data platforms and engineering
  • Building a custom MCP for your own data and tools
  • AI-enabling your data team or development team with Cursor

Email svm@mitc.dk, call +45 23 61 21 91, or find Svend on LinkedIn. All three are on the contact page.

You can also ask the server itself. The contact_mitc tool returns the same details.

Contact