EUVD-MCP

MCP server for the European Union Vulnerability Database (EUVD) maintained by ENISA.

Features

• Search vulnerabilities with flexible filters (CVSS, EPSS, dates, product, vendor, exploited status, etc.)

• Get latest, critical, and exploited vulnerabilities

• Lookup specific vulnerabilities and advisories by ID

• Automatic retries with exponential backoff

• Bounded TTL cache (cachetools.TTLCache) for list endpoints

• Structured logging (always to stderr — safe for stdio transport)

• /health liveness endpoint and /metrics observability endpoint (HTTP mode)

• Startup connectivity check to the EUVD API

• Two transport modes: HTTP (standalone/Docker) and stdio (subprocess/Claude Desktop)

Related MCP server: vuln-nist-mcp-server

Requirements

• Python 3.14+

• Poetry

Installation

git clone <repository-url>
cd euvdmcp
poetry install

Copy the example environment file and adjust as needed:

cp .env.template .env

Configuration

All settings are read from environment variables (or a .env file at the project root).

Running

uvicorn (recommended)

uvicorn euvd_mcp.main:app --host 127.0.0.1 --port 8000

With live reload during development:

uvicorn euvd_mcp.main:app --host 127.0.0.1 --port 8000 --reload

Poetry

poetry run python -m euvd_mcp.main

Server runs on http://127.0.0.1:8000/mcp by default.

Docker

Build the image:

make docker-build

Run with a .env file (configuration is not baked into the image):

make docker-run        # uses .env automatically

Or docker-compose:

make compose-up        # start
make compose-logs      # tail logs
make compose-down      # stop

stdio (Claude Desktop / subprocess)

TRANSPORT=stdio poetry run python -m euvd_mcp.main

Note: In stdio mode the /health and /metrics endpoints are not available. Logs go to stderr so they do not interfere with the MCP protocol on stdout.

Integrating with LLM Clients

Claude Desktop — HTTP (server running separately)

Add to your Claude Desktop configuration file:

macOS/Linux: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "euvd": {
      "url": "http://127.0.0.1:8000/mcp"
    }
  }
}

Claude Desktop — stdio (subprocess)

{
  "mcpServers": {
    "euvd": {
      "command": "poetry",
      "args": ["run", "python", "-m", "euvd_mcp.main"],
      "env": { "TRANSPORT": "stdio" }
    }
  }
}

Claude Desktop — Docker

{
  "mcpServers": {
    "euvd": {
      "command": "docker",
      "args": [
        "run", "--rm", "-p", "8000:8000",
        "--env-file", "/absolute/path/to/.env",
        "euvd-mcp:local"
      ]
    }
  }
}

Example queries

• "What are the latest critical vulnerabilities?"

• "Search for exploited vulnerabilities with CVSS score above 8.0"

• "Get details for vulnerability EUVD-2024-45012"

• "Find vulnerabilities from Microsoft published in the last 30 days"

Available Tools

Observability

Two endpoints are available in HTTP mode:

Project Structure

euvdmcp/
├── euvd_mcp/
│   ├── main.py                    # MCP server, tool definitions, health/metrics routes
│   ├── controllers/
│   │   └── euvd_api.py            # Async API client with retry and TTL cache
│   ├── models/
│   │   ├── input_models.py        # Pydantic input validation models
│   │   └── vulnerability.py       # Response data models
│   ├── utils/
│   │   ├── settings.py            # Configuration (pydantic-settings)
│   │   ├── logging_config.py      # Structured logging setup
│   │   └── metrics.py             # In-process metrics collector
│   └── tests/                     # pytest test suite
├── Dockerfile                     # Multi-stage build, non-root user, HEALTHCHECK
├── docker-compose.yml
├── Makefile
└── pyproject.toml

Development

poetry install                 # install all dependencies (incl. dev)
make pre-commit-setup          # install git hooks
make test                      # run tests
make test-cov                  # run tests with coverage report
make lint                      # ruff + mypy
make format                    # auto-format with ruff
make security                  # bandit + pip-audit

CI

Three GitHub Actions workflows run on every PR to main:

• CI — tests, security scan (bandit + pip-audit), lock-file check, Docker build

• Code Quality — ruff lint/format, mypy type check, markdown validation

• Security Scan — Trivy container scan, results uploaded to GitHub Security tab

License

See LICENSE

Author

Duarte Dias

Acknowledgments

• ENISA — European Union Agency for Cybersecurity

• FastMCP — MCP framework