Skip to main content
Glama
datakoot

Security Intel MCP

README.md
# Security Intel MCP — by Datakoot

Vulnerability intelligence for AI agents — as MCP tools your agent can call mid-task. One keyless call fuses NVD, CISA KEV and FIRST EPSS. No API keys.

## Tools

| Tool | What it does | Source |
|---|---|---|
| `cve_lookup` | CVE summary: description, CVSS score & severity, CWE, references, plus whether it is actively exploited (CISA KEV) and its exploit probability (EPSS) | NVD (NIST), CISA KEV, FIRST EPSS |
| `known_exploited` | Is a CVE on CISA's Known Exploited Vulnerabilities catalog? Or list the newest exploited CVEs (filter by vendor, product or ransomware use) | CISA KEV |
| `epss_score` | Exploit probability (0–1) and percentile for one or many CVEs: the chance each is exploited in the next 30 days | FIRST EPSS |
| `package_vulnerabilities` | Known vulnerabilities for a package/version | OSV.dev |
| `audit_dependencies` | Audit a whole package.json (or dependency list) in one call | OSV.dev |

No API keys required for any tool.

## Quick start

```
claude mcp add --transport http security-intel https://security.datakoot.com/mcp
```

Or point any MCP client at `https://security.datakoot.com/mcp`.

## Try it in 10 seconds — no key, no signup

Paste this into a terminal:

```bash
curl -s https://security.datakoot.com/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "cve_lookup", "arguments": {"cve_id": "CVE-2021-44228"}}}'
```

You get Log4Shell (CVE-2021-44228) in one answer: the NVD record (severity, CVSS vector, references), its CISA KEV entry (exploited in the wild, used in ransomware) and its EPSS exploit probability. No API key, nothing to sign up for.

Or point any MCP client at the URL and just ask your agent, in plain language:

- "Is CVE-2021-44228 something I need to worry about?"
- "Which of these CVEs is actually being exploited right now?"
- "Audit my package.json for known vulnerabilities before I deploy."


## Data & attribution

Vulnerability data comes from the [National Vulnerability Database](https://nvd.nist.gov) (NIST — US public domain), CISA's [Known Exploited Vulnerabilities catalog](https://www.cisa.gov/known-exploited-vulnerabilities-catalog) (US public domain), [FIRST EPSS](https://www.first.org/epss/) and [OSV.dev](https://osv.dev) (CC-BY 4.0), the same open sources used by scanners like Trivy and Grype.

Part of [Datakoot](https://datakoot.com) — keyless intelligence APIs for AI agents.

TDQS

A3.8/5.0

Scored across 5 tools

Disambiguation4/5

The five tools mostly target distinct resources: individual CVE lookup, KEV catalog queries, EPSS scoring, per-package OSV lookup, and manifest-wide auditing. However, cve_lookup already bundles KEV and EPSS results inline, so an agent may be unsure when to call the dedicated known_exploited or epss_score tools versus the composite lookup. The package_vulnerabilities vs audit_dependencies split is clear (single package vs whole manifest).

Naming Consistency3/5

Names are readable but follow mixed conventions: cve_lookup, known_exploited, epss_score, package_vulnerabilities, and audit_dependencies blend noun_noun, adjective-style, and verb_noun forms. Two tools (audit_dependencies, package_vulnerabilities) use verb-based action naming while others are noun phrases. Not chaotic, but no single predictable pattern.

Tool Count5/5

Five tools is a tight, well-scoped surface for a security-intelligence server covering CVE data, exploit likelihood, and dependency auditing. Each tool earns its place without redundancy or thinness. No unused or filler tools.

Completeness4/5

The surface covers the core workflows: CVE detail, known-exploited status, EPSS prioritization, package lookup, and full manifest auditing. Minor gaps exist, such as keyword/vendor-based CVE search or advisory text search, but agents can work around these via CVE IDs and OSV queries.

Maintenance

ActivityActive
ResponsivenessNo issues