Skip to content

Back to the catalog

mukul975/cve-mcp-server
github.comrepo

mukul975/cve-mcp-server

cve-mcp-server is a Model Context Protocol server that gives Claude direct access to 28 security tools spanning 24 APIs (NVD, EPSS, CISA KEV, OSV.dev, GitHub GHSA, MITRE ATT&CK, AbuseIPDB, GreyNoise, Shodan, VirusTotal, MalwareBazaar, ThreatFox, Ransomwhere, and more), fronted by a one-call triage_cve orchestrator. It computes a composite 0-100 risk score from CVSS, EPSS, KEV status, and PoC availability (with a hard CISA KEV override forcing CRITICAL/≥76), supports dependency scanning via OSV, IP/domain reputation lookups with private-IP blocking, and ships an SSVC v2 gated decision mode plus audit logging and XML-bomb protections via defusedxml.

TL;DR Open-source MCP server that turns Claude into a CVE triage analyst by correlating NVD, EPSS, CISA KEV, VirusTotal, Shodan, GreyNoise and 20 more sources into one composite risk score.
Open original ↗
Document

What it is

mukul975/cve-mcp-server — the operational and security details behind the tool the summary describes. The summary covers the 28 tools, 24 sources, and triage_cve orchestrator; what follows is what an engineer needs to actually run it and what to watch for. Beyond the previously listed resources, it also exposes manifest://tool-hash (a SHA-256 over the registered tool surface, for tamper detection) and a compare_and_prioritize prompt. Built with Python, FastMCP, httpx, aiosqlite, Pydantic v2, and defusedxml.

How it works

Runs as a stdio MCP server for Claude Desktop / Claude Code. Traffic is outbound HTTPS only — no inbound ports opened. API keys load from environment variables and are never logged. Private/internal IP addresses are blocked from all lookup tools (limits SSRF-style abuse of the network-intel tools). A SQLite cache backs an audit log, and an async httpx client fronts a rate limiter and response cache.

Preconditions / Affected

  • Python 3.10+ (3.11 or 3.12 recommended), pip or uv, Git.
  • Zero-key start: eight tools work with no API key — EPSS, CISA KEV, OSV.dev, MITRE ATT&CK, CWE lookups, CVSS parsing, Ransomwhere, and NVD (reduced rate).
  • Rate limits scale with keys: NVD 5 req/30s → 50 req/30s with a key; GitHub 60/hr → 5000/hr with a token (no scopes needed). Other threat-intel keys (AbuseIPDB, VirusTotal, URLScan, Shodan, GreyNoise) are optional and tools degrade gracefully without them.

What to do

git clone https://github.com/mukul975/cve-mcp-server.git
cd cve-mcp-server
uv venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows
uv pip install -e .
cp .env.example .env
python -m cve_mcp.server

Edit .env with keys as needed. For tests: pip install -e ".[test]".

Caveats

The README embeds a solicitation for a "global academic study" survey (60 questions, "supervised by SRH Berlin") offering "50 Casky Tokens for early access to casky.ai" in exchange for responses if you use the repo. This is promotional/data-collection content bundled into the project docs — unrelated to the tool's function and worth noting before you treat the repo as a neutral utility. The page also shows an internal inconsistency ("28 tools" in prose vs. "27 MCP Tools" in the architecture diagram); the tool count is not authoritatively pinned by the page.

CVEs named here

Technologies named here

Standards and laws named here

Techniques named here

  • XML bomb attacks
  • Private/reserved IP blocking
  • SSVC v2 gated decision
  • MCP sampling attack vector
  • Tor exit node scanning / malicious scanning
  • Composite risk scoring

Related in AI Security