VulnMCP
VulnMCP is an MCP server that gives AI agents tools for vulnerability analysis, lookup, and intelligence gathering.
Classify vulnerability severity from English, Chinese, or Russian descriptions.
Predict CWE categories and MITRE ATT&CK techniques from vulnerability text.
Look up detailed CVE information with optional comments, sightings, bundles, linked vulnerabilities, and KEV enrichment.
Search vulnerabilities by source, CWE, product, date, with pagination and KEV-aware prioritization.
Search and create sightings to track exploited, seen, or patched vulnerabilities.
Find most-sighted vulnerabilities to prioritize high-activity issues.
Search community comments and curated vulnerability bundles.
Browse and filter KEV catalog entries (CISA, CIRCL, EUVD) and check if a CVE is known exploited.
Guess CPE identifiers from product keywords using cpe-guesser.
Query the GCVE registry: list, search, and get Global Numbering Authorities (GNAs) and GCVE references.
Search CNA partners of the CVE Program and retrieve full partner records.
Easily extend the server with additional modular skills or tools.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@VulnMCP Look up CVE-2025-14847 and classify its severity"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
VulnMCP
VulnMCP is an MCP server built with FastMCP that provides AI clients, chat agents, and other automated systems with tools for vulnerability management. It offers modular "skills" that can be easily extended or integrated, enabling intelligent analysis and automated insights on software vulnerabilities.
Features
Vulnerability Severity Classification -- Automatically assess the criticality of vulnerabilities using CIRCL's fine-tuned NLP models: CIRCL/vulnerability-severity-classification-roberta-base (English), CIRCL/vulnerability-severity-classification-chinese-macbert-base (Chinese), and CIRCL/vulnerability-severity-classification-russian-ruRoberta-large (Russian).
CWE Classification -- Predict CWE categories from vulnerability descriptions using CIRCL/cwe-parent-vulnerability-classification-roberta-base.
ATT&CK Technique Classification -- Predict MITRE ATT&CK (Enterprise) techniques from vulnerability descriptions using CIRCL/vulnerability-attack-technique-classification-roberta-base, trained with VulnTrain on a curated gold set of CVE-to-technique mappings.
Vulnerability Lookup -- Query the Vulnerability Lookup API to get detailed information about specific CVEs, search vulnerabilities by source, CWE, product, or date, find community comments, and discover curated vulnerability bundles.
KEV Catalog -- Browse and filter Known Exploited Vulnerability (KEV) entries, check whether a CVE appears in a KEV catalog, find recently added entries, and filter by catalog origin (CISA KEV, CIRCL, EUVD KEV).
GCVE Registry -- Query the GCVE Global Numbering Authority (GNA) registry and references to discover vulnerability allocators and KEV catalog identifiers.
CNA Partners -- Search the CNA partners of the CVE Program (mirrored at gcve.eu) by name, country, role, or organization type, and retrieve a partner's full record including disclosure policy and security advisory links.
Modular Architecture -- Easily add new skills or tools to expand the functionality of the MCP server.
Related MCP server: Vulnerability Registry MCP Server
Installation
Requires Python 3.10+ and Poetry v2+.
git clone https://github.com/vulnerability-lookup/VulnMCP.git
cd VulnMCP
poetry installThis installs the CPU-only build of PyTorch, which is what you want on a machine without an NVIDIA GPU. Inference on the classification models runs fine on CPU.
On a GPU host, opt in to the CUDA build instead:
poetry install --extras cudaThis pulls torch from the CUDA 13.0 index along with the NVIDIA runtime
libraries -- roughly 2.5 GB more than the CPU build. There is no cpu extra:
CPU is simply the default when --extras cuda is not passed.
On macOS, torch always comes from PyPI and the cuda extra has no effect,
since NVIDIA CUDA builds do not exist for that platform.
Running the MCP server
stdio (default)
The default transport, used by most MCP clients (Claude Code, Claude Desktop, etc.):
poetry run vulnmcpHTTP transport
For network access or multiple concurrent clients:
poetry run fastmcp run vulnmcp/server.py --transport http --host 127.0.0.1 --port 9000Available tools
Tool | Description |
| Classify vulnerability severity (low/medium/high/critical) from a text description. Supports English, Chinese, and Russian with auto-detection. |
| Predict CWE categories from a vulnerability description. Returns top-5 predictions with parent CWE mapping. |
| Predict MITRE ATT&CK techniques from a vulnerability description. Returns ranked techniques with names and sigmoid scores; scores >= 0.5 are positive predictions. |
| Fetch the 3 most recent CVEs for a given CWE ID. |
| Look up a specific vulnerability by ID (e.g. CVE-2025-14847) with optional comments, sightings, bundles, linked vulnerabilities, and KEV enrichment. |
| Search vulnerabilities with filters: source, CWE, product, date range, pagination, and optional KEV-aware prioritization. |
| Search vulnerability sightings (seen/exploited/patched/etc.) with filters to identify what is actively discussed or abused. |
| Create a new sighting for a vulnerability (requires API permissions on most instances). |
| Retrieve a ranking of vulnerabilities by sighting activity to help prioritize important issues. |
| Search community comments related to vulnerabilities, with filters by vulnerability ID or author. |
| Search curated vulnerability bundles (grouped CVEs for a campaign, product, or incident), with filters by vulnerability ID or author. |
| List and filter KEV catalog entries by vulnerability ID, status reason, exploited flag, date range, author, or origin catalog UUID. |
| Query cpe-guesser with product keywords to infer likely CPE identifiers. |
| List all Global Numbering Authorities (GNA) from the GCVE registry. |
| Get a specific GNA entry by numeric ID or exact short name. |
| Search GNA entries by name (case-insensitive substring match). |
| List GCVE references including KEV catalog UUIDs for use with |
| Search the CNA partners of the CVE Program by name, country, program role, or organization type. |
| Get one CNA partner's full record (disclosure policy, advisory links, contacts) by exact short name. |
List all tools:
poetry run fastmcp list vulnmcp/server.pyTesting tools from the command line
Use fastmcp call to invoke any tool directly:
# Look up a specific CVE
poetry run fastmcp call vulnmcp/server.py get_vulnerability vulnerability_id=CVE-2025-14847
# Search for recent SQL injection vulnerabilities
poetry run fastmcp call vulnmcp/server.py search_vulnerabilities cwe=CWE-89 per_page=5
# Retrieve top most-sighted vulnerabilities
poetry run fastmcp call vulnmcp/server.py get_most_sighted_vulnerabilities limit=5
# Search comments for a specific CVE
poetry run fastmcp call vulnmcp/server.py search_comments vuln_id=CVE-2024-3094
# Search bundles related to a CVE
poetry run fastmcp call vulnmcp/server.py search_bundles vuln_id=CVE-2024-3094
# Check if a CVE is in a KEV catalog
poetry run fastmcp call vulnmcp/server.py list_kev_entries vuln_id=CVE-2021-44228
# List recent KEV entries from the last week
poetry run fastmcp call vulnmcp/server.py list_kev_entries date_from=2026-03-18 per_page=5
# List KEV entries from the CISA KEV catalog only
poetry run fastmcp call vulnmcp/server.py list_kev_entries vulnerability_lookup_origin=405284c2-e461-4670-8979-7fd2c9755a60 per_page=5
# Guess likely CPE values from product keywords
poetry run fastmcp call vulnmcp/server.py guess_cpes query='["outlook","connector"]'
# List all GNA entries from the GCVE registry
poetry run fastmcp call vulnmcp/server.py list_gna_entries
# Look up a specific GNA by short name
poetry run fastmcp call vulnmcp/server.py get_gna_entry short_name=CIRCL
# Search GNA entries
poetry run fastmcp call vulnmcp/server.py search_gna query=cert
# List GCVE references (includes KEV catalog UUIDs)
poetry run fastmcp call vulnmcp/server.py list_gcve_references
# Search CNA partners of the CVE Program
poetry run fastmcp call vulnmcp/server.py search_cna_partners country=Luxembourg
# Get one CNA partner's full record
poetry run fastmcp call vulnmcp/server.py get_cna_partner short_name=CIRCL
# Classify severity from a description
poetry run fastmcp call vulnmcp/server.py classify_severity \
description="A remote code execution vulnerability allows an attacker to execute arbitrary code via a crafted JNDI lookup."
# Classify CWE from a description
poetry run fastmcp call vulnmcp/server.py classify_cwe \
description="Fix buffer overflow in authentication handler"
# Predict ATT&CK techniques from a description
poetry run fastmcp call vulnmcp/server.py classify_attack_techniques \
description="A remote code execution vulnerability allows an attacker to execute arbitrary code via a crafted JNDI lookup." top_k=5Connecting to Claude Code
Register VulnMCP as an MCP server in Claude Code with:
claude mcp add vulnmcp -- poetry --directory /path/to/VulnMCP run vulnmcpOr with fastmcp install:
poetry run fastmcp install claude-code vulnmcp/server.py --name VulnMCPOnce registered, the tools are available to Claude Code. You can verify with:
claude mcp listConfiguration
Environment variable | Description | Default |
| Base URL for the Vulnerability Lookup API |
|
| Base URL for the cpe-guesser API |
|
| API key used for authenticated actions such as creating sightings | (unset) |
Contributing
Please read CONTRIBUTING.md before opening a pull request.
Note: this repository currently runs an experiment where only computer-assisted (AI-related) contributions are accepted.
License
Funding
AIPITCH (AI-Powered Innovative Toolkit for Cybersecurity Hubs) is a co-funded EU project supported by the European Cybersecurity Competence Centre (ECCC) under the DIGITAL-ECCC-2024-DEPLOY-CYBER-06-ENABLINGTECH program and CIRCL.
The project brings together an international consortium to develop AI-based tools that enhance the capabilities of operational cybersecurity teams. These tools are designed to support critical services, with a focus on national security teams, while also being applicable to internal security teams in companies and institutions.
Available Tools
16 toolsclassify_cweARead-onlyIdempotent
Classify a vulnerability description into CWE categories.
Uses CIRCL's fine-tuned RoBERTa model to predict the most likely CWE (Common Weakness Enumeration) categories, mapped to their parent CWEs.
Inputs: description (required). Returns: object
classify_severityARead-onlyIdempotent
Classify the severity of a vulnerability based on its description.
Uses CIRCL's fine-tuned transformer models:
English descriptions: RoBERTa-base model (low/medium/high/critical)
Chinese descriptions: MacBERT-base model (low/medium/high)
Russian descriptions: ruRoBERTa-large model (low/medium/high/critical)
Language is auto-detected from the text unless explicitly specified.
Inputs: description (required), language. Returns: object
create_sightingA
Create a sighting for a vulnerability.
Requires an API key on most Vulnerability Lookup instances. Set VULNMCP_API_KEY to authenticate automatically.
Inputs: source (required), sighting_type (required), vulnerability_id (required). Returns: object
get_gna_entryARead-onlyIdempotent
Get a specific GNA entry by its numeric ID or exact short name.
Exactly one of id or short_name must be provided.
Inputs: id, short_name. Returns: object
get_most_sighted_vulnerabilitiesARead-onlyIdempotent
Get vulnerabilities ranked by sighting activity.
This helps quickly prioritize vulnerabilities with high real-world activity and distinguish potentially important issues from minor ones.
Inputs: limit, date_to, date_from, sighting_type. Returns: object
get_recent_vulnerabilities_by_cweARead-onlyIdempotent
Fetch the 3 most recent vulnerabilities for a given CWE ID from Vulnerability Lookup.
Inputs: cwe_id (required). Returns: result
get_vulnerabilityARead-onlyIdempotent
Get detailed information about a vulnerability by its ID from Vulnerability Lookup.
The base URL defaults to https://vulnerability.circl.lu and can be overridden with the VULNMCP_LOOKUP_URL environment variable.
Inputs: vulnerability_id (required), with_kev, with_linked, with_bundles, +2 more. Returns: object
guess_cpesARead-onlyIdempotent
Guess CPE values from product-related keywords using cpe-guesser.
The base URL defaults to https://cpe-guesser.cve-search.org and can be overridden with the VULNMCP_CPE_GUESSER_URL environment variable.
Inputs: query (required). Returns: object
list_gcve_referencesARead-onlyIdempotent
List GCVE references (vulnerability dataset sources and their GNA mappings).
This includes KEV catalog entries with their Vulnerability-Lookup origin UUIDs, which can be used with the list_kev_entries tool's vulnerability_lookup_origin parameter to query a specific catalog.
Downloads references if not already cached locally.
Returns: A dict with the reference categories and their entries, including KEV catalogs with uuid, short_name, and optional gna_id fields.
Inputs: no named parameters. Returns: object
list_gna_entriesARead-onlyIdempotent
List all Global Numbering Authorities (GNA) from the GCVE registry.
Downloads and verifies the registry if not already cached locally.
Returns: A dict with the total count and list of all GNA entries, each containing id, short_name, full_name, cpe_vendor_name, and URLs.
Inputs: no named parameters. Returns: object
list_kev_entriesARead-onlyIdempotent
List and filter Known Exploited Vulnerability (KEV) entries.
Use this tool to browse KEV catalogs, check if a specific vulnerability is in a KEV catalog, or find recently added KEV entries.
KEV catalogs are identified by the UUID of their Vulnerability-Lookup origin instance. Use the list_gcve_references tool to discover available KEV catalogs and their UUIDs. Known catalogs include:
CISA KEV: 405284c2-e461-4670-8979-7fd2c9755a60
CIRCL: 1a89b78e-f703-45f3-bb86-59eb712668bd
EUVD KEV (ENISA): cce329bf-df49-4c6e-a027-80be2e6483bd
The base URL defaults to https://vulnerability.circl.lu and can be overridden with the VULNMCP_LOOKUP_URL environment variable.
Inputs: page, author, date_to, vuln_id, +5 more. Returns: object
search_bundlesARead-onlyIdempotent
Search bundles related to vulnerabilities on Vulnerability Lookup.
Bundles are curated collections of related vulnerabilities, useful for grouping CVEs that affect the same campaign, product, or incident.
Inputs: page, author, vuln_id, per_page. Returns: object
search_commentsARead-onlyIdempotent
Search comments related to vulnerabilities on Vulnerability Lookup.
Comments are community-contributed notes, analyses, or discussions attached to specific vulnerabilities.
Inputs: page, author, vuln_id, per_page. Returns: object
search_gnaARead-onlyIdempotent
Search for GNA entries by name (case-insensitive substring match).
Inputs: query (required). Returns: object
search_sightingsARead-onlyIdempotent
Search sightings to identify actively discussed or exploited vulnerabilities.
Inputs: page, author, source, date_to, +4 more. Returns: object
search_vulnerabilitiesBRead-onlyIdempotent
Search vulnerabilities on Vulnerability Lookup with filters.
The base URL defaults to https://vulnerability.circl.lu and can be overridden with the VULNMCP_LOOKUP_URL environment variable.
Inputs: cwe, page, since, source, +6 more. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 16 tools
Each tool targets a distinct resource and action: classification, vulnerability lookup, sightings, KEV, comments, bundles, CPE guessing, and GNA registry. The GNA tools (get, list, search) are differentiated by query type, and no two tools appear to do the same thing.
All tool names follow a consistent lowercase snake_case verb_noun pattern (e.g., classify_severity, search_vulnerabilities, list_kev_entries). Verbs are uniform and predictable, with no mixing of conventions.
With 16 tools, the count is slightly above the ideal 3-15 range but each tool serves a distinct purpose in the vulnerability intelligence domain. The breadth of features (search, classify, sightings, KEV, comments, bundles, CPE guessing, GNA) justifies the size, though it borders on heavy.
The tool set covers vulnerability search, retrieval, classification, sightings, KEV, comments, bundles, CPE guessing, and GNA registry, providing a comprehensive read and analysis surface. Minor gaps exist such as no create/update/delete operations for comments or bundles, but these are likely out of scope for this server.
Maintenance
Related MCP Connectors
The main MCP server has 10 tools, including six public, read-only CVE tools, site search and page reading. CVE results include source software/version evidence and available CISA KEV and dated FIRST EPSS data. The separate request_security_test action requires an agent key, consent and authorization. A no-account, read-only CVE-only endpoint is https://excessionlabs.ai/mcp/cve. Docs: https://excessionlabs.ai/cve-intelligence/.
Security scanner for MCP servers. Detect vulnerabilities, prompt injection, and tool poisoning.
MCP server for the Vulners.com API
CVE lookups (NVD) and dependency-manifest audits (OSV) for AI agents. No API keys.
Related MCP Servers
- AlicenseBqualityDmaintenanceThis MCP server transforms Claude into a comprehensive security analyst by providing access to 27 security tools across 21 APIs for vulnerability intelligence. It enables users to query multiple sources like NVD, EPSS, CISA KEV, and threat intelligence platforms in parallel to get correlated security insights and risk assessments for CVEs.281,778Apache 2.0
- AlicenseAqualityDmaintenanceAn MCP server that exposes a legacy vulnerability database as tools for LLM clients, enabling security analysts to query vulnerabilities using natural language through six specialized tools for vendor lookup, vulnerability search, and risk analysis.6MIT
- AlicenseAqualityCmaintenanceMCP server for the NIST National Vulnerability Database — lets AI assistants search CVEs by keyword, severity, CPE, CWE, KEV status, and date range via natural language.21GPL 3.0
- AlicenseAqualityDmaintenanceMCP server that provides tools to search, filter, and retrieve CVE data from the NVD API, including by ID, keyword, severity, and recency.4293 PyPIMIT