CVE Hunter MCP Agent

A robust Model Context Protocol (MCP) server built to empower LLMs with real-time cybersecurity intelligence. This agent interfaces with the CIRCL.lu API to fetch, analyze, and summarize Common Vulnerabilities and Exposures (CVEs) without requiring expensive API keys.

Designed with FastMCP, Pydantic validation, and asynchronous architecture.

---

Features

• Intelligent Search: Drills down from Vendor -> Product -> Vulnerabilities (e.g., "Find bugs in Microsoft" -> "List Office bugs").
• CISO-Level Briefings: specialized prompts to generate executive summaries for specific CVEs.
• Real-time Intelligence: Fetches the latest published vulnerabilities globally.
• Robust Validation: Uses Pydantic to enforce valid CVE ID formats and sanitize inputs before API calls.

---

Quick Start

Prerequisites

• Python 3.12+
• uv (Fast Python package installer)
• Google Gemini API Key (For the demo client)

Option 1: Running with Docker (Recommended)

Spin up the server instantly without installing Python dependencies locally.

docker compose up --build

The server will start on http://0.0.0.0:8000.

Option 2: Running Locally (Development)

Install Dependencies

pip install uv
uv sync

Start the Server

uv run src/server/main.py

Start the Client (New Terminal) Make sure you export your API key first:

export GOOGLE_API_KEY="your-gemini-api-key"
uv run src/client/client.py

💡 Example Usage

Once the client is running, you can interact with the agent using natural language:

• Deep Dive - "Act as a security researcher and analyze CVE-2021-44228."
• Discovery - "Does Microsoft Office have any recent vulnerabilities?"
• Monitoring - "What are the top 5 latest CVEs published today?"
• Validation - "Check CVE-2024-9999" (Agent will validate ID format)