A zero-dependency Model Context Protocol server that bridges stdio-only MCP clients to the hosted NotCVE MCP server for vulnerability intelligence across 365k+ CVEs/NotCVEs from 80+ correlated sources.
17 tools with complete schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools are READ_ONLY (good for safety), but descriptions lack actionable guidance for LLM selection. Parameter descriptions are absent, LLMs cannot infer what 'vulnerability_id' vs 'cve_id' vs 'notcve_id' mean or when to use each. No output schemas documented. Naming is verb-forward (get_, list_, search_) but some tools conflate similar concepts (get_cve vs get_notcve vs get_vulnerability_summary). Error handling and recovery guidance missing entirely.
Your plan, remaining credits and derived capability flags. Free (0 credits).
Summaries for up to 10 CVE/NotCVE ids in one round-trip. 1 credit.
Yes/no impact of one CVE/NotCVE for a concrete vendor/product/version. 1 credit.
Matrix of which CVEs affect which versions of one vendor/product. 1 credit.
Full canonical record for one CVE id (metadata + CPEs + TIR stubs; no TIR bodies). 1 credit.
One-call exploit/patch material for a CVE: clean held bodies + reference URLs. 1 credit + 1 per extra held body.
No parameter descriptions. LLMs cannot infer what 'vulnerability_id' means or when to use get_cve vs get_notcve vs get_vulnerability_summary. Ambiguous parameter names force guessing.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. Responses are opaque, agents cannot chain tools effectively.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | <=2025-11-25 | v2 |
Full canonical record for one NotCVE id. 1 credit.
Compact CVE/NotCVE summary plus retrieval advice. 1 credit.
One TIR content blob — clean payload (rule body, exploit code, patch) + mime/len + download URL. 1 credit.
Affected-version rows for one CVE/NotCVE, paginated 50/page. 1 credit.
Paginated TIR metadata for one CVE/NotCVE; filter by type / source. 1 credit.
Recently added CISA KEV entries. 1 credit.
Reference for the search filters and modes. Free (0 credits).
Typed-argument search across 365k+ CVEs/NotCVEs (query, vendor, product, version, CVSS, exploited, date ranges). search_mode=fast costs 1 credit, deep costs 2.
Triage up to 10 stack components (vendor/product/version each) at once. 1 credit.
Aggregate Threat-Intelligence-Resource counts by type and by source for one CVE/NotCVE. Free (0 credits).
CVEs affecting a given vendor/product/version, with a severity rollup. 1 credit.
Confusing tool naming: get_cve, get_notcve, get_vulnerability_summary all retrieve a single record but use different parameter names (cve_id, notcve_id, vulnerability_id). LLMs cannot distinguish when to use each.
No error handling or recovery guidance. If a tool fails, LLMs have no actionable next step. No indication of retryable vs fatal errors.
Pagination not documented. list_vulnerability_tirs and list_affected_versions mention pagination but no limit/offset/cursor parameters visible in schema. LLMs cannot control result size.