NVD Database
A Model Context Protocol server implementation to query the NIST National Vulnerability Database (NVD) via its API. https://nvd.nist.gov/
As a prerequisite an NVD API key is required. (Request here).
Status
Works with Claude Desktop app and other MCP compliant hosts and clients using both the stdio and sse transports.
Features
- Query specific CVEs by ID with detailed vulnerability data.
- Search the NVD database by keyword with customizable result options.
- Supports Server-Sent Events (SSE) transport for real-time communication.
- Compatible with MCP-compliant clients like Claude Desktop.
Tools
The server implements the following tools to query the NVD Database:
-
get_cve:- Description: Retrieves a CVE record by its ID.
- Parameters:
cve_id(str): The CVE ID (e.g.,CVE-2019-1010218).concise(bool, defaultFalse): IfTrue, returns a shorter format.
- Returns: Detailed CVE info including scores, weaknesses, and references.
-
search_cve:- Description: Searches the NVD database by keyword.
- Parameters:
keyword(str): Search term (e.g.,Red Hat).exact_match(bool, defaultFalse): IfTrue, requires an exact phrase match.concise(bool, defaultFalse): IfTrue, returns shorter CVE records.results(int, default10): Maximum number of CVE records (1-2000).
- Returns: List of matching CVEs with total count.
Configuration
-
Create or edit the Claude Desktop configuration file located at:
- On macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - On Windows:
%APPDATA%/Claude/claude_desktop_config.json
- On macOS:
-
Add the following:
{
"mcpServers": {
"mcp-nvd": {
"command": "/path/to/uvx",
"args": ["mcp-nvd"],
"env": {
"NVD_API_KEY": "your-api-key"
}
}
}
}
-
Replace
/path/to/uvxwith the absolute path to theuvxexecutable. Find the path withwhich uvxcommand in a terminal. This ensures that the correct version ofuvxis used when starting the server. -
Restart Claude Desktop to apply the changes.
Development
Setup
-
Prerequisites:
- Python 3.10 or higher.
- An NVD API key (request here).
uvpackage manager (installation).
-
Clone the Repository:
git clone https://github.com/marcoeg/mcp-nvd
cd mcp-nvd
-
Set Environment Variables:
- Create a
.envfile in the project root:NVD_API_KEY=your-api-key - Replace
your-api-keywith your NVD API key.
- Create a
-
Install Dependencies:
uv sync
uv pip install -e .
Run with the MCP Inspector
cd /path/to/the/repo
source .env
npx @modelcontextprotocol/inspector uv \
--directory /path/to/repo/mcp-nvd run mcp-nvd
Then open the browser to the URL indicated by the MCP Inspector, typically http://localhost:8077?proxyPort=8078
Switch freely between
stdioandssetransport types in the inspector.
Testing with the SSE Client
Run the Server:
cd /path/to/the/repo
source .env
uv run mcp-nvd --transport sse --port 9090
- Runs with SSE transport on port
9090by default.
Run the Client:
Test get_cve:
uv run client.py http://localhost:9090/sse CVE-2019-1010218
Test search_cve (default 10 results):
uv run client.py http://localhost:9090/sse "search:Red Hat"
Test search_cve (exact match, 5 results):
uv run client.py http://localhost:9090/sse "search:Microsoft Windows:exact:5"
Docker Setup
Build
docker build -t mcp-nvd:latest .
Run
With .env:
docker run -d -p 9090:9090 -v /path/to/.env:/app/.env mcp-nvd:latest
With env var:
docker run -d -p 9090:9090 -e NVD_API_KEY="your-key" mcp-nvd:latest
Custom port:
docker run -d -p 8080:8080 -v /path/to/.env:/app/.env mcp-nvd:latest uv run mcp-nvd --transport sse --port 8080 --host 0.0.0.0
Verify
docker logs <container_id>
# Expect: INFO: Uvicorn running on http://0.0.0.0:9090
Test:
uv run client.py http://localhost:9090/sse CVE-2019-1010218
Notes
- Ensure
.envhasNVD_API_KEY=your-keyor use-e. - Default port:
9090.
Here’s the summary formatted as Markdown comments within a code block, suitable for inclusion in a file like docker-compose.yaml or README.md:
Using Docker Compose for Testing
This docker-compose.yaml, located in the tests/ directory, defines a service for testing the MCP-NVD server using a pre-built Docker image. It’s designed for a testing use case, similar to a standalone service like clickhouse, and assumes the image is built beforehand rather than rebuilt each time.
Assumptions
- Pre-built Image: The service uses a pre-built image tagged as
mcp-nvd:test, available locally or in a registry. The image is based on theDockerfilein the parent directory, which sets up the MCP-NVD server withuvand runs it in SSE mode on port 9090.
How to Build the Image
To create the mcp-nvd:test image:
- Navigate to the project root:
cd ./mcp-nvd - Build the image using the Dockerfile:
docker build -t mcp-nvd:test .- This builds the image with all dependencies from
pyproject.tomland themcp_nvd/module, setting the default command to run the server.
- This builds the image with all dependencies from
Running the Service
From the tests/ directory:
cd tests
docker-compose up
- Access: The server runs at
http://localhost:9090. - Stop:
docker-compose down. - Environment: Ensure
NVD_API_KEYis in../.envor usedocker-compose --env-file ../.env up.
Running test_tools.py in the Docker Compose Scenario
To run the unit tests (test_tools.py) within the Docker environment:
- Start the Service: Ensure the
mcp-nvdservice is running viadocker-compose up. - Exec into the Container:
- Identify the container name (e.g.,
mcp-nvd-mcp-nvd-1) with:docker ps - Run the tests inside the container:
docker exec -it mcp-nvd-mcp-nvd-1 python /app/tests/test_tools.py - Note: Assumes
test_tools.pyis copied into the image at/app/tests/. If not, modify the Dockerfile to include:Then rebuild the image withCOPY tests/ ./tests/docker build -t mcp-nvd:test .from the root.
- Identify the container name (e.g.,
- Alternative: Run tests locally against the containerized service:
cd tests python test_tools.py- This tests against
http://localhost:9090while the service runs.
- This tests against
Key Details
- Port: 9090 is exposed for SSE access.
- Logs: Stored in a
log-datavolume (optional). - Image: Must be built once and tagged as
mcp-nvd:testbefore runningdocker-compose.
Credits to @sidharthrajaram for its working pattern for SSE-based MCP clients and servers: https://github.com/sidharthrajaram/mcp-sse