Aleva data in your AI assistant

Connect

The server speaks Streamable HTTP at:

https://api.aleva.se/mcp

You need no account and no key. Add the address as an MCP server or custom connector in your client. Everything is read-only: the tools never change anything.

Clients

Claude

In the Claude app, add a custom connector with the address above. In Claude Code:

claude mcp add --transport http aleva https://api.aleva.se/mcp

ChatGPT

Add a custom connector or app with the address above (developer mode needs to be on in ChatGPT settings).

Cursor

~/.cursor/mcp.json

{
  "mcpServers": {
    "aleva": {
      "url": "https://api.aleva.se/mcp"
    }
  }
}

VS Code

.vscode/mcp.json

{
  "servers": {
    "aleva": {
      "type": "http",
      "url": "https://api.aleva.se/mcp"
    }
  }
}

Other clients work the same way: all they need is the address.

Limits and keys

A tool call counts as a normal API call to the same endpoint, with the same limits. Without a key: 10 cost units a minute and 200 a day per IP address. With a key, the key's limits apply. Send it as a header in the client's configuration:

{
  "mcpServers": {
    "aleva": {
      "url": "https://api.aleva.se/mcp",
      "headers": {
        "Authorization": "Bearer alv_…"
      }
    }
  }
}

Responses carry their source and licence. When you show data from a tool, credit the source as you would for the API. More about responses.

Tools

Two tools help the assistant find its way: search_catalogue searches the datasets, and describe_endpoint describes the fields, licence and freshness of one. Then there is one tool per endpoint: