Jeevesus — DugganUSA Threat-Intelligence MCP Server

A read-only research desk for threat intel, inside your AI workflow.

Live now. Listed on the official MCP Registry as io.github.pduggusa/dugganusa-threat-intel since 2026-05-02.

Endpoint: https://analytics.dugganusa.com/api/v1/mcp

What It Does

Jeevesus exposes the DugganUSA threat-intelligence platform as a Model Context Protocol server, so Claude Desktop, Cursor, Continue, and any other MCP-compatible client can query it without leaving the editor or chat.

You get instant access to:

1.12M+ IOCs across 37 Meilisearch indexes (IPs, domains, URLs, hashes)
1,555 CISA KEV CVEs with patch status and exploitation timelines
366 adversaries profiled with TTPs and infrastructure
16.8K OTX pulses correlated with the cascade pipeline
400K+ Epstein documents for OSINT research
Full STIX 2.1 feed corpus — same data your SIEM consumes

No tab-switching out of your AI workflow.

Tools

Jeevesus ships with a static, allow-listed tool set. No dynamic tool registration, no tool-poisoning surface.

`search`

Full-text search across 9 allow-listed indexes: iocs, cves, adversaries, otx_pulses, stix_objects, mcp_servers, mcp_findings, whitepapers, epstein_docs.

{
  "name": "search",
  "arguments": {
    "index": "iocs",
    "query": "185.220.101.0/24",
    "limit": 10
  }
}

Indexes outside the allow-list are rejected at the server boundary — clients cannot pivot to internal-only indexes.

`enrich-ioc`

Enrich a single IOC (IP, domain, URL, or SHA256 hash) with cross-source correlation: VirusTotal, AbuseIPDB, ThreatFox, our cascade pipeline, and behavioral scoring.

{
  "name": "enrich-ioc",
  "arguments": {
    "ioc": "185.220.101.42",
    "type": "ip"
  }
}

Cached for 24 hours per IOC across all callers.

`stix-feed-summary`

Returns a daily summary of the STIX 2.1 feed: counts by indicator type, top adversaries, new IOCs in the last 24 hours, and feed-health metrics.

Auth required — this tool is gated behind a Bearer key on every tier.

{
  "name": "stix-feed-summary",
  "arguments": { "since": "2026-05-01T00:00:00Z" }
}

Authentication

Jeevesus uses the same API keys as the rest of the DugganUSA platform.

Authorization: Bearer <your_api_key>

Important: Use the Authorization header. The X-API-Key header is stripped at the edge and will not authenticate.

Tier	Price	Daily Limit	Per-Minute	SLA
Free	$0	25/day	10/min	Community
Pro	$99/mo ($948/yr)	2,000/day	60/min	24h email
Enterprise	$995/mo ($9,950/yr)	50,000/day	unlimited	4h response

Limits are shared across the STIX feed, REST API, and MCP server — one key, one quota.

Install

Claude Desktop

Add the following to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "dugganusa-threat-intel": {
      "url": "https://analytics.dugganusa.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Restart Claude Desktop. The three tools appear in the tool tray.

Cursor

Open Settings → MCP → Add new MCP server and paste the same config. Cursor reloads tools automatically.

Continue

In ~/.continue/config.json:

{
  "mcpServers": [
    {
      "name": "dugganusa-threat-intel",
      "url": "https://analytics.dugganusa.com/api/v1/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  ]
}

Registry install

Anywhere mcp-publisher is supported:

mcp-publisher install io.github.pduggusa/dugganusa-threat-intel

The CLI prompts for your API key and writes the appropriate client config.

Security Posture

Jeevesus is read-only and secure-by-default. Concretely:

Static tool list. No tools/list mutation between sessions; clients see exactly three tools, always.
Allow-listed indexes. search enforces a server-side allow-list of 9 indexes — requests for anything else are rejected, not just filtered.
Prompt-injection sanitization. Tool outputs are passed through a sanitizer that strips system-style markers, model-control prefixes, and known jailbreak patterns before reaching the client.
Rate limiting. Per-key and per-IP. Same limits as the REST API.
Hashed-IP audit log. Every call is logged with a salted hash of the client IP — never the raw address. Rotated salts make logs unjoinable across windows.
No write tools. Jeevesus cannot create, modify, or delete anything. There is no upload, no submit, no create_* surface.
No shell, no file, no network egress tools. Even a fully compromised key can only read.
TLS-only. Plaintext requests are rejected at the edge.

We hold ourselves to the same standard we audit other servers against — see the AIPM-for-MCP audit methodology.

Examples

Triage an alerting IP

“Is 185.220.101.42 worth blocking? Show me what we know.”

Claude calls enrich-ioc and returns a synthesized answer with VT/AbuseIPDB/ThreatFox scores, ASN/ownership, behavior tags, and the cascade pipeline confidence — in one turn.

CVE patch prioritization

“Of the CVEs in our last scan, which are in CISA KEV and have public exploit code?”

Claude calls search against the cves index with the scan list as filter terms, intersects with KEV-tagged entries, and ranks by exploitation date.

Pre-merge supply-chain check

“Run the SHA256 of this binary against the IOC index before I merge.”

enrich-ioc returns hash matches across malware families and known campaigns. If clean, you merge. If hot, you don’t.

Daily standup

“What changed in the STIX feed in the last 24 hours?”

stix-feed-summary produces the diff. Paste it into Slack.

Security.DugganUSA.com Documentation