@qune-tech/ocds-mcp

MCP server for German public procurement data (OCDS). Connects your AI assistant to the Vergabe Dashboard API for semantic search, tender matching, and company profile management.

Company profiles never leave your machine. GDPR-compliant by design.

Quick Start

npx @qune-tech/ocds-mcp --api-key sk_live_YOUR_KEY_HERE

Get an API key

Sign up at vergabe-dashboard.qune.de and create an API key (MCP or Enterprise plan required).

Configure your AI client

Claude Desktop

Edit claude_desktop_config.json:

{
  "mcpServers": {
    "ocds": {
      "command": "npx",
      "args": ["-y", "@qune-tech/ocds-mcp", "--api-key", "sk_live_YOUR_KEY_HERE"]
    }
  }
}

Claude Code

Add .mcp.json to your project root:

{
  "mcpServers": {
    "ocds": {
      "command": "npx",
      "args": ["-y", "@qune-tech/ocds-mcp", "--api-key", "sk_live_YOUR_KEY_HERE"]
    }
  }
}

Cursor

Settings > MCP Servers > Add:

• Command: npx
• Args: -y @qune-tech/ocds-mcp --api-key sk_live_YOUR_KEY_HERE

Available Tools

| Tool | Description | |------|-------------| | search_text | Semantic search across all tenders | | list_releases | Filter and browse tenders by month, CPV code, category, value range | | get_release | Full tender details by OCID | | get_index_info | Database statistics and connectivity check | | create_company_profile | Create a matching profile for your company | | update_company_profile | Update an existing profile | | get_company_profile | View profile details | | list_company_profiles | List all your profiles | | delete_company_profile | Delete a profile | | match_tenders | Match a profile against all tenders with semantic similarity |

How It Works

The npm package downloads the correct platform-native binary on install. No Node.js runtime dependency for the actual MCP server.

LLM <--stdio--> ocds-mcp (local binary)
                   |  Local: company profiles + sentence embedder
                   |  Remote: searches, release queries
                   +--HTTPS--> Vergabe Dashboard API

• Company profiles are stored locally (never leave your machine).
• Text embeddings are computed locally (multilingual-e5-small ONNX model, ~118 MB, auto-downloaded on first run).
• Only embedding vectors are sent to the API for search and matching.

Supported Platforms

| Platform | Architecture | |----------|-------------| | Linux | x86_64 | | macOS | Apple Silicon (ARM64) | | Windows | x86_64 |

Requirements

• An API key from vergabe-dashboard.qune.de
• ~200 MB disk space for the ONNX model (auto-downloaded on first run)
• Internet connection to reach the API

License

MIT