CVE MCP Server

# CVE Loading Benchmark Results **Date**: November 26, 2025 **Test Environment**: Ubuntu Linux, Docker ## Executive Summary **Winner: Git Clone Method** - 342x faster than GitHub API once repository is cloned. ## Methods Tested 1. **GitHub API (Individual Files)**: Current implementation 2. **Git Clone (Local Parsing)**: New optimized method ## Results ### Speed Comparison (500 CVEs) | Method | Time | Speed | Winner | |--------|------|-------|--------| | GitHub API | 62.5 sec | 8 CVEs/sec | | | Git Clone | 0.18 sec | 2,735 CVEs/sec | | **Speed Improvement: 342x faster** ### Full Dataset Projections (240,000 CVEs) | Method | Estimated Time | Estimated Size | |--------|---------------|----------------| | GitHub API | 8.3 hours | 156 MB | | Git Clone | 1.5 minutes | 189 MB | ### Initial Setup Cost - **Git Clone**: 5 minutes (one-time repository clone) - **GitHub API**: None ### Space Efficiency - **GitHub API**: 680 bytes/CVE - **Git Clone**: 827 bytes/CVE (+22% larger, negligible difference) ## Recommendation **Use Git Clone method for:** - Initial database population - Full dataset loads - Bulk updates **Total time for 240K CVEs:** - Clone repo: ~5 minutes - Load all CVEs: ~1.5 minutes - **Total: ~6.5 minutes** vs 8.3 hours with API **Final database size: ~190 MB** (acceptable) ## Raw Benchmark Data [See full benchmark output in benchmarks/ directory]