sanctum-mcp

JasonTofte/sanctum-mcp

★ 0 stars Python AI/LLM Updated today

Architecturally-hardened DFIR MCP server — SANS FIND EVIL! hackathon submission. Autonomous incident response agent with architectural evidence-integrity and prompt-injection defense.

View on GitHub → Try with Claude — $10 free →

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "sanctum-mcp": {
      "command": "uvx",
      "args": [
        "sanctum-mcp"
      ]
    }
  }
}

Or install with pip: pip install sanctum-mcp

README Excerpt

> **Results (DFIR-Metric subset, 43 questions × 3 runs, Claude Opus 4.7):** > Sanctum **99.2%** [95.7, 99.9] vs. a bare model **16.3%** [10.9, 23.6] — an **82.9-point gap** (Wilson 95% CIs). Precision on CORROBORATED findings **97.2%**; false-confidence rate **2.8%**. Method: [`docs/ACCURACY.md`](docs/ACCURACY.md).

sanctum-mcp

Quick Install

README Excerpt

Related AI/LLM Servers