Guide
How to look up CVEs in ChatGPT with live NVD data
ChatGPT does not know recent CVEs and will guess. Here is how to give it live CVE data from NVD, CISA KEV, and EPSS over MCP, so it answers from the source instead.
Add the CVE brain to your AI assistant
Ask ChatGPT for the details of a recent CVE and you will often get a confident answer that is out of date or simply invented. That is not a flaw you can prompt your way around: the model only knows what it was trained on, and CVEs are published every day. The fix is to give ChatGPT a live source it can query at the moment you ask. The KBrain CVE brain does exactly that over MCP, so ChatGPT returns CVSS scores, CISA KEV status, and EPSS probabilities from data refreshed daily.
Why ChatGPT gets CVEs wrong on its own
- Training cutoff: any CVE published after the model's cutoff is unknown to it, so it fills the gap with a plausible guess.
- No freshness signal: a CVSS score can be revised and a CVE can be added to CISA KEV after publication, and the model cannot see those changes.
- High stakes: in security work, a wrong exploitation status or severity is worse than no answer.
The fix: connect a live CVE brain over MCP
The KBrain CVE brain is a hosted knowledge source that ChatGPT can call while it answers. It exposes tools to look up a CVE by id, search by keyword or product, list recent CVEs, and check CISA KEV status. Because it is hosted and refreshed daily, there is nothing for you to run and no API key to manage.
- get_cve: full detail for one CVE, including CVSS, affected products, references, KEV status, and EPSS.
- search_cves: keyword or product search with severity and date filters.
- get_recent_cves: what was published in the last N days, optionally by minimum severity.
- get_kev_status: a quick exploited yes or no with the date added.
How to connect it to ChatGPT
- Add the CVE brain to your KBrain account, then open your MCP connector settings.
- Connect the KBrain endpoint to ChatGPT, following the standard MCP setup for ChatGPT.
- Ask a CVE question. ChatGPT calls the brain, retrieves the current record, and answers from it.
For the exact connection steps, see the guide on connecting a knowledge base to ChatGPT via MCP. Once connected, the CVE tools appear automatically alongside your other brains.
Questions to try
- "Give me the full details for CVE-2024-3400."
- "What critical CVEs were published this week?"
- "Search for recent OpenSSL vulnerabilities rated high or above."
- "Is CVE-2023-46805 being actively exploited?"
Every answer is grounded in the current record and carries a last-updated timestamp, so you can tell when the data was refreshed rather than trusting a number the model recalled.
Add the CVE brain to your AI assistant
Look up CVEs, CISA KEV exploited status, and EPSS scores from Claude, ChatGPT, or any MCP compatible assistant. Hosted and refreshed daily, with no self-hosting and no API keys to manage.
Frequently asked questions
Why does ChatGPT get recent CVEs wrong?
ChatGPT answers from training data with a fixed cutoff. Any CVE published after that cutoff is unknown, so the model guesses. CVSS revisions and new CISA KEV entries after publication are also invisible to it. Connecting a live CVE source over MCP replaces the guess with a lookup.
Do I need an NVD API key to use this in ChatGPT?
No. The KBrain CVE brain is hosted and handles the data collection for you. You connect the brain to ChatGPT once; there is no key to request or rotate and no server to run.
Does this work with ChatGPT Actions?
Yes. KBrain generates an OpenAPI specification for ChatGPT in addition to the MCP tool schema, so the CVE brain works as a ChatGPT Action as well as through MCP.
How current is the CVE data?
CVE records sync from NVD daily, and CISA KEV and EPSS are refreshed daily as well. Every answer carries a last-updated timestamp so you can confirm freshness.