# vulntel

> Vulnerability intelligence your AI agent can trust. A hosted MCP server that fuses NVD, CISA KEV, FIRST EPSS, OSV/GHSA and CISA SSVC into exploitation-first vulnerability intelligence for AI agents, and fact-checks the CVEs an agent might hallucinate.

## Start here

- [Home](https://vulntel.com/): Vulnerability intelligence your AI agent can trust.
- [Tools reference](https://vulntel.com/tools): What each of the 13 tools returns, with live examples.
- [New in bug bounty scope](https://vulntel.com/scope): Weekly: assets that newly entered public bug-bounty scope on HackerOne, Bugcrowd, Intigriti and YesWeHack, re-adds and renames filtered out (RSS: https://vulntel.com/scope/feed.xml).

## Guides & docs

- [Vulnerability Intelligence APIs Compared (2026): vulntel vs VulnCheck vs Vulners vs OpenCVE](https://vulntel.com/vulnerability-intelligence-api-comparison): An honest 2026 comparison of vulnerability intelligence APIs: which fuse NVD, CISA KEV, FIRST EPSS, OSV/GHSA and SSVC, which are MCP-native for AI agents, and where each one is genuinely stronger. Facts dated, corrections welcome.
- [The Best Vulnerability Intelligence API for CVE Enrichment (2026)](https://vulntel.com/best-vulnerability-intelligence-api): There is no single best vulnerability intelligence API; the right one depends on your job. This page names the best pick for AI-agent triage, for raw source breadth, for first-party exploit data, and for a free self-hosted option, with the reasoning behind each.
- [MCP Servers for AI Bug-Bounty Agents, Compared (November 2026)](https://vulntel.com/mcp-servers-for-bug-bounty-agents): Which MCP servers and agent tools help an AI bug-bounty agent: CVE intelligence, disclosed-report mechanics, program priors, fresh scope, distro-aware banner checks. What each does, what it costs, whether it is hosted, and where each one is stronger. Every claim sourced and dated.
- [Vulnerability Intelligence MCP for AI Agents](https://vulntel.com/vulnerability-intelligence-mcp): vulntel is a Model Context Protocol (MCP) server that gives an AI agent source-grounded vulnerability intelligence: NVD, CISA KEV, FIRST EPSS, OSV/GHSA and CISA SSVC fused into exploitation-first answers, plus a fact-check for the CVEs an agent might hallucinate.
- [What Is CVE Enrichment?](https://vulntel.com/guides/what-is-cve-enrichment): CVE enrichment takes a bare CVE identifier and attaches the context needed to act on it: severity, known-exploited status, exploitation probability, affected products and versions, patches, and public exploit code. A raw CVE ID tells you almost nothing; an enriched CVE tells you whether to drop everything and patch.
- [What Is EPSS Exploit Prediction?](https://vulntel.com/guides/what-is-epss): EPSS (the Exploit Prediction Scoring System, from FIRST) is a 0-to-1 probability that a CVE will be exploited in the wild within 30 days. Unlike CVSS, which scores severity, EPSS scores likelihood of attack, so it tells you what to fix first.
- [EPSS API Documentation](https://vulntel.com/docs/epss): How to get FIRST EPSS scores programmatically with vulntel: the enrich_cve MCP tool returns a CVE's EPSS score and percentile fused with KEV, SSVC and a PoC signal. Includes the response shape and how to reach FIRST's own EPSS API.
- [What Is a Vulnerability Intelligence API?](https://vulntel.com/guides/what-is-a-vulnerability-intelligence-api): A vulnerability intelligence API is a programmatic endpoint that returns enriched, prioritized CVE data: not just the raw record, but exploitation status, likelihood, affected versions, and exploit availability, so software or an AI agent can decide what to fix first.
- [How to Use CVE Enrichment for Bug Bounty Triage](https://vulntel.com/guides/cve-enrichment-for-bug-bounty): A step-by-step workflow for using CVE enrichment on a bug bounty engagement: catalog the stack, enrich each finding with KEV/EPSS/SSVC, verify the claim, and rank by exploitation-first priority so you spend recon time where a bug is most likely to be real.

## Field notes

- [Does an agent skill change how an AI hunts? We A/B tested our own](https://vulntel.com/agent-skill-exhortation-vs-enforcement): We ran three identical bug-hunting agents on one target, changing only the preamble: no skill, our 400-line skill, and a rewritten enforced version. The current skill barely changed the agent's behavior. Enforcement did. Here is the data, the code, and the honest ceiling that stopped us from claiming more.
- [Vulnerability intelligence for AI agents: the complete guide](https://vulntel.com/ai-agent-vulnerability-intelligence): The complete, data-backed guide to giving an AI security agent real vulnerability intelligence: why a CVE lookup is not enough, how to prioritize by exploitation (EPSS, KEV, SSVC over CVSS) with real numbers from 368,000 CVEs, why agents hallucinate CVEs and how to catch it, transferring attack mechanics across products, the honest ceiling on autonomous agents, and the tool workflow that ties it together.
- [The bug-hunting skill: teaching an AI agent to think like an attacker, not an engineer](https://vulntel.com/ai-agent-bug-hunting-skill): Why we distilled a bug-bounty skill from real engagement logs instead of reaching for a bigger model: the generative half that sees the non-obvious bug, the disciplinary half that keeps a clever idea from becoming a fabricated report, and why both are required.
- [From twelve tools to a hunt: the playbook for driving a vulnerability-intelligence MCP](https://vulntel.com/bug-bounty-mcp-playbook): How we came up with the MCP playbook: the per-tool value map that separates the differentiated tools from the commodity lookups, the operating loop in order, and the disciplines you only learn by losing time (browser UA, read-the-fix-to-falsify, transfer-and-execute, verify-before-claim).
- [You can't turn a pull-based MCP into a push model (it's a memory problem, not a prompt one)](https://vulntel.com/pull-vs-push-mcp): Why an MCP server can't make an agent follow a multi-step loop by nudging it from tool responses, why 'prompt injection for good' both breaks trust and backfires, and why the push you actually want turns out to be persistent state, not clever phrasing.
- [How to catch AI-hallucinated CVEs before they reach a report](https://vulntel.com/catch-hallucinated-cves): AI agents invent CVE IDs that do not exist, assign the wrong severity, and attribute a bug to the wrong product. Here is how to fact-check every CVE claim against a ground-truth corpus (NVD + OSV + GHSA + CISA KEV) in one call, with real refuted examples.
- [The bottleneck for AI security agents isn't reasoning (it's memory and discipline)](https://vulntel.com/ai-security-agent-bottleneck): Field notes from pointing AI agents at real bug-bounty targets: the ceiling is an environment boundary, not model quality; the highest-value memory is your failures, which no CVE database contains; and why a confident claim is a hypothesis to falsify, not a result.
- [Do software vulnerabilities transfer between products? Measuring mechanism reuse across 14,000 disclosed bugs](https://vulntel.com/vulnerability-mechanism-transfer): A data analysis of ~10,000 disclosed HackerOne reports distilled into product-agnostic 'mechanism cards' and embedded: the median vulnerability mechanic has already been independently filed against 4 different products, a third against 11+, and 27% of decade-old mechanics still recruit a brand-new product every 18 months. Vulnerability mechanisms are reused - and keep spreading - far more than the product-centric (one-CVE-per-product) view suggests. With method, numbers, and limitations.
- [What actually predicts vulnerability exploitation? Lessons from 330,000 CVEs](https://vulntel.com/predicting-exploitation): A data analysis of 330k CVEs with CISA KEV as ground truth: EPSS vs CVSS, the exploit-PoC reference as a near-perfect filter, exploitation magnets by CWE and ecosystem, and how time-to-exploit collapsed to days.
- [Can a CVE corpus make an LLM a better bug hunter? Eight experiments](https://vulntel.com/llm-bug-hunting-experiments): We tested whether a 330k-CVE corpus turns an LLM into a better vulnerability researcher, mechanism pattern-cards, invariant reconstruction, cheap-model uplift, change-history prediction, and vulnerable-vs-patched discrimination. Honest results, including the failures.
- [Vulnerability prioritization that works: SSVC, EPSS, and KEV over CVSS](https://vulntel.com/vulnerability-prioritization): Why patching by CVSS fails, what a defensible P1–P4 priority model looks like, and how CISA's SSVC reconstructs the KEV exploited-list at 99.9%, with worked examples (BlackLotus, the xz backdoor, Log4Shell).

## Sources

- [NVD](https://nvd.nist.gov/): Authoritative source fused into the corpus.
- [CISA KEV](https://www.cisa.gov/known-exploited-vulnerabilities-catalog): Authoritative source fused into the corpus.
- [FIRST EPSS](https://www.first.org/epss/): Authoritative source fused into the corpus.
- [OSV](https://osv.dev/): Authoritative source fused into the corpus.
- [GHSA](https://github.com/advisories): Authoritative source fused into the corpus.
- [CISA Vulnrichment](https://github.com/cisagov/vulnrichment): Authoritative source fused into the corpus.
- [Debian Security Tracker](https://security-tracker.debian.org/tracker/): Authoritative source fused into the corpus.
- [Ubuntu CVE Tracker](https://ubuntu.com/security/cves): Authoritative source fused into the corpus.
- [Alpine secdb](https://secdb.alpinelinux.org/): Authoritative source fused into the corpus.