Apiverket MCP Server

An MCP (Model Context Protocol) server that gives AI assistants access to Swedish public data through the Apiverket API. It works with Claude Code, Claude Desktop, Cursor, VS Code, Gemini CLI-style MCP clients, and other MCP-compatible tools.

What it does

Agents should call govdata_discover before govdata_query instead of guessing /v1 paths. After a 429 or when a user asks about limits, agents should call govdata_account.

Related MCP server: Kolada MCP Server

How Agents Should Use Apiverket

1. Discover: call govdata_discover with the user's topic, not a guessed path.

2. Check context: call govdata_account for quota, tier, and production-readiness questions.

3. Query: call govdata_query only with a supported endpoint returned by discovery.

4. Recover: if a request fails, use the structured recovery guidance before retrying.

Discovery includes family workflows for transport, weather, geography, municipalities, jobs, government, culture, education, tourism, environment, and company data. These workflows explain search/list/detail patterns, required codes, coordinates, date formats, pagination, and common recovery steps.

Quick Start

Claude Code / Claude Desktop

{
  "mcpServers": {
    "apiverket": {
      "command": "npx",
      "args": ["-y", "apiverket-mcp-server"],
      "env": {
        "GOVDATA_API_URL": "https://apiverket.se",
        "GOVDATA_API_KEY": "sk_test_demo"
      }
    }
  }
}

Cursor / VS Code

{
  "servers": {
    "apiverket": {
      "command": "npx",
      "args": ["-y", "apiverket-mcp-server"],
      "env": {
        "GOVDATA_API_URL": "https://apiverket.se",
        "GOVDATA_API_KEY": "sk_test_demo"
      }
    }
  }
}

Gemini CLI-Style MCP Config

{
  "mcpServers": {
    "apiverket": {
      "command": "npx",
      "args": ["-y", "apiverket-mcp-server"],
      "env": {
        "GOVDATA_API_URL": "https://apiverket.se",
        "GOVDATA_API_KEY": "sk_test_demo"
      }
    }
  }
}

Keys And Limits

sk_test_demo is for setup and safe sandbox responses. It is useful for checking that your MCP client is wired correctly, but production workflows should use a live key. Free live keys are good for exploration; upgrade when daily API or company-search limits block a real workflow.

Company Data With Agents

Company search and company lookup have different jobs:

1. Use govdata_discover(query: "company") to find the supported company endpoints.

2. Use /v1/companies/search when the user only has a company name or uncertain organisation number.

3. Cache the returned org_number.

4. Use /v1/companies/{orgNumber} for repeated enrichment and automation.

5. If company search returns 429, stop retrying until reset_at; use lookup when org numbers are already known.

Company search and live company lookup have separate daily quota context by tier. govdata_account shows the configured key tier, remaining company-search quota, remaining company-lookup quota, reset time, and available upgrade tiers. Apiverket does not expose company board, officer, owner, or UBO subresource paths through the company API.

Common Family Workflows

Tool Examples

# Discover endpoints
govdata_discover(query: "electricity prices")
govdata_discover(category: "Weather & Climate")
govdata_discover(query: "train departures")
govdata_discover(query: "municipality KPI")
govdata_discover(query: "jobs taxonomy")

# Query data after discovery
govdata_query(endpoint: "/v1/weather/{city}", path_params: {city: "stockholm"})
govdata_query(endpoint: "/v1/transport/trains/{station}", path_params: {station: "Cst"})
govdata_query(endpoint: "/v1/jobs/search", query_params: {q: "developer", limit: 5})
govdata_query(endpoint: "/v1/taxonomy/types")
govdata_query(endpoint: "/v1/companies/search", query_params: {q: "volvo", limit: 10})
govdata_query(endpoint: "/v1/companies/{orgNumber}", path_params: {orgNumber: "5560125790"})

# Inspect tier and quota state
govdata_account()

Build From Source

git clone https://github.com/vinvuk/apiverket-mcp.git
cd apiverket-mcp
npm install
npm run build

How It Works

1. The MCP server runs locally over stdio.

2. govdata_discover searches the built-in endpoint catalog without an API call.

3. govdata_query calls Apiverket with the configured API key.

4. Structured API errors are turned into recovery guidance for agents.

5. Large responses are truncated with pagination guidance.

Requirements

• Node.js >= 18

• An Apiverket API key, or sk_test_demo for sandbox setup

License

MIT