Kolada

Kolada MCP Server provides a remote MCP endpoint and local installation to expose Kolada's municipal and regional statistics to AI copilots and LLMs. It exposes a range of tools (KPI, municipalities, organizational units, data access, and analysis) that you can query via MCP endpoints or through the supported client transports (HTTP/SSE) depending on your integration. This server enables remote access to 5,000+ KPI records across 264 operating areas for all Swedish municipalities and regions, making it practical to pull metadata, perform searches, and retrieve time-series or filtered KPI data for dynamic AI-assisted decision support.

You can connect from clients like Claude, Claude Desktop, OpenAI Codex CLI, Gemini, and others using the provided URLs. The server supports multiple transport modes (remote HTTP/SSE for remote clients, and local stdio with npx). Typical usage involves pointing your client to the MCP endpoint at /mcp (or /sse for Server-Sent Events) and then invoking the available tools, such as search_kpis, get_kpi, or get_kpi_data, depending on your needs. The API is organized into modules for KPI data, municipalities/regions, organizational units, data retrieval, and analysis, enabling both simple lookups and complex filtering or comparisons across municipalities.

For example, you can: (1) search for KPI metadata with search_kpis, (2) fetch a single KPI’s metadata via get_kpi, (3) retrieve time-series or filtered KPI data with get_kpi_data, or (4) compare municipalities or KPI groups to support analytics within your AI assistant. The server is designed to be robust for AI-assisted workflows, with caching, rate limiting, and multilingual Swedish documentation for ease of use in local contexts.

Prerequisites:

• Node.js and npm installed on your machine (or access to a host that can run npx).
• Internet access to fetch kolada-mcp-server from npm registry (when using npx).

Option A: Quick start with npx (no local install)

1. Ensure Node.js is installed.
2. Run the MCP server with npx:

npx -y kolada-mcp-server

3. The server will start and listen on the default port (usually 3000) or as configured by environment variables. Use the provided /mcp endpoint for clients.

Option B: Local installation (persistent server)

1. Install globally (optional, for persistent running):

npm install -g kolada-mcp-server

2. Start the server locally:

kolada-mcp-server

3. Connect clients to http://localhost:3000/mcp (or your configured host/port).

Option C: Use in a container (docker)** This project can be run via npm/pnpm/yarn in your environment or via npx as shown above. If you prefer a container workflow, you can build a small image that runs npm install kolada-mcp-server and then launches kolada-mcp-server. See your preferred container tooling for running a Node.js app.

Tips and common issues:

• Endpoint URLs: /mcp for standard MCP transport, /sse for Server-Sent Events. Use /health for health checks.
• If you run behind a proxy or in a restricted network, ensure the port used by the MCP server is accessible and properly forwarded.
• When using CLI clients (Claude, Codex, Gemini), point them at the exact mcp URL and transport type (HTTP/SSE) as documented for each client.
• Environment variables can be used to configure host/port, rate limits, and cache behavior; consider setting MCPP_PORT, MCPP_HOST, and related settings if you deploy behind reverse proxies.
• If you see rate-limit or timeout errors, verify network connectivity and consider increasing timeouts in client configurations or enabling server-side rate limiting tuning if supported by your deployment.
• Regularly update to the latest kolada-mcp-server version to benefit from security patches and new tools/support.