MCP Spotlight: WhisperGraph — 46 Billion Nodes of Internet Intelligence, Queryable by Your Agent

Server: WhisperGraph by whisper-sec Stars: 376 · License: Apache 2.0 · Language: TypeScript MCP Transport: stdio + Streamable HTTP + hosted remote (mcp.whisper.security) Graph scale: 46B nodes/edges, 20 entity types, ~40 threat feeds · Last updated: May 17, 2026

What It Does

Internet infrastructure data is scattered across a dozen services. DNS records sit in one place, BGP routing history in another, WHOIS registrations in a third, and threat intelligence feeds in a fourth. To answer a single question — "is this domain malicious?" — an analyst chains together six different queries across four different tools.

WhisperGraph unifies all of it into a single Cypher-queryable graph database: 46 billion nodes and edges mapping every domain, IP, ASN, prefix, organization, web link, and threat-intelligence listing into one connected model. DNS resolution, BGP origin history, WHOIS contacts, SPF chains, and ~40 threat feeds across 18 categories — all reachable through the same MCP server.

And it's built for agents from day one. Signup is two HTTP calls — no browser, no CAPTCHA. An AI agent can provision its own API key programmatically in ~5 seconds.

Why It Matters for Agent Engineering

Security operations is one of the most natural domains for AI agents — and one where tool fragmentation causes the most friction. The typical investigation workflow looks like this:

1. Check DNS records → dig or a DNS API
2. Look up BGP origin → BGP.tools or RIPEstat API
3. Pull WHOIS registrant → a WHOIS API
4. Check threat feeds → VirusTotal, AbuseIPDB, etc.
5. Trace related infrastructure → repeat steps 1–4 for every related domain/IP

Each step involves a different tool, a different API key, a different response format. The agent spends most of its time on orchestration, not investigation.

WhisperGraph collapses this into a single step:

MATCH (d:Domain {name:'suspicious.example.com'})-[:RESOLVES_TO]->(ip:IP)-[:HOSTED_BY]->(asn:ASN)
MATCH (ip)-[:LISTED_ON]->(ti:ThreatIntel)
RETURN d, ip, asn, ti

One Cypher query. DNS, BGP, and threat intel in a single response. The agent can follow relationships across entity types without changing tools, API endpoints, or response parsers.

The MCP Tool Surface

WhisperGraph exposes six read-only tools, designed for the investigation workflow:

Tool	Purpose
query	Execute Cypher queries against the 46B-node graph
list_labels	List all 20 node labels with counts — anchors your query
describe_label	Confirm a label exists and enumerate its property keys
explain_indicator	Threat assessment for IP, hostname, CIDR, or ASN — score, level, factors
whisper_history	Historical WHOIS or BGP data for an indicator
domain_variants	Typosquatting and brand-protection variants, checked against the graph

All queries go through a safety rule set before reaching the backend. The MCP server validates Cypher syntax and rejects destructive operations at the protocol level — your agent can't accidentally DELETE from the graph.

Eight Investigation Prompts, Built In

One of WhisperGraph's smartest design choices: it ships with eight MCP prompt templates that give the LLM a structured starting point for common investigations:

Prompt	What It Guides the Agent To Do
investigate-ip	Full threat assessment: DNS, BGP, WHOIS, threat feeds
map-attack-surface	Enumerate all domains, IPs, and ASNs for an organization
compare-domains	Side-by-side comparison of two domains' infrastructure
blast-radius	Trace all infrastructure connected to a compromised indicator
threat-triage	Quick scoring and prioritization of a list of indicators
whois-pivot	Find all domains registered by the same entity
bgp-investigation	Route hijacking and MOAS conflict analysis
typosquat-sweep	Discover registered lookalike domains targeting a brand

These aren't just documentation — they're executable workflows the agent can invoke directly. An incident responder can say "do a blast radius analysis on this IP" and the agent follows the structured investigation template without needing step-by-step instruction.

Agent-First Design

WhisperGraph made a deliberate architectural choice that sets it apart from most API-based MCP servers: programmatic agent signup.

Most SaaS APIs require a human to visit a website, fill out a form, solve a CAPTCHA, and copy an API key. WhisperGraph's signup is two API calls:

POST https://console.whisper.security/api/signup
Content-Type: application/json

{"email":"your-agent@example.com","attribution":{"agent_name":"my-agent","source":"facio"}}

Whisper emails a verification code. The agent retrieves it and calls:

POST https://console.whisper.security/api/signup/verify

{"signup_id":"...","code":"..."}

The response contains api_key, mcp_url, dashboard_url, and docs_url. Working API key in ~5 seconds, entirely automated. This is how all SaaS APIs should work for agent workflows — and almost none of them do.

Connecting WhisperGraph to Facio

Step 1: Agent Signup (Programmatic)

Your Facio agent can provision its own WhisperGraph API key. The two-call signup flow means no manual browser steps — the agent registers, verifies email, and receives credentials autonomously.

Step 2: Use the Hosted Remote Server

No installation needed. Whisper runs a hosted MCP server at https://mcp.whisper.security:

{
  "mcpServers": {
    "whisper": {
      "url": "https://mcp.whisper.security",
      "headers": {
        "Authorization": "Bearer ${credentials.WHISPERGRAPH_API_KEY}"
      }
    }
  }
}

Step 3: Or Self-Host the MCP Layer

For teams that want the MCP server in their own environment:

docker run -p 8080:8080 -e MCP_TRANSPORT=http \
  -e WHISPER_API_KEY=${credentials.WHISPERGRAPH_API_KEY} \
  ghcr.io/whisper-sec/whisper-graph-mcp:latest

The self-hosted Docker container relays queries to the WhisperGraph API — your API key stays in your environment, and you control the network path.

Production Patterns

Automated Threat Triage

With WhisperGraph connected to Facio, a security agent can run autonomous triage on alert feeds:

Agent workflow (triggered by SIEM alert):
1. Extract IP addresses and domains from alert payload
2. For each indicator: whisper.explain_indicator → threat score + factors
3. For high-severity indicators: whisper.query to trace blast radius
4. Aggregate into a triage report with prioritized response actions
5. Route high-risk findings through Facio's HITL approval for escalation

The agent handles the repetitive triage work — checking 40 threat feeds, tracing infrastructure connections, scoring indicators. Human analysts review only the high-severity cases that need decision-making.

Brand Protection Monitoring

The domain_variants tool enables autonomous typosquatting detection:

1. Agent maintains a list of monitored domains (your brands, your clients)
2. Cron-triggered: whisper.domain_variants for each domain
3. Cross-reference results against the graph to check if variants are registered and hosted
4. Alert on new registrations that match known malicious infrastructure patterns
5. Generate takedown recommendations with evidence

All automated. The agent doesn't just detect — it traces the infrastructure to build a complete picture for takedown requests.

Incident Response Blast Radius

When a compromise is confirmed, the first question is always: "what else is affected?"

MATCH (compromised:IP {address:'203.0.113.42'})-[:HOSTED_BY]->(asn:ASN)
MATCH (asn)<-[:HOSTED_BY]-(sibling:IP)-[:RESOLVES_TO]-(domain:Domain)
MATCH (sibling)-[:LISTED_ON]->(ti:ThreatIntel)
RETURN domain.name, sibling.address, ti.feed, ti.category

One query traces from a compromised IP to its ASN, finds all sibling IPs in the same ASN, resolves their domains, and checks threat intelligence listings. What would take a human analyst 30 minutes of tab-switching takes the agent 30 seconds.

Data Breadth

WhisperGraph's 46 billion nodes cover a remarkable range:

Category	Coverage
DNS	Resolution, nameservers, MX, SPF chains, DNSSEC
BGP	ASN ownership, origin history, MOAS conflicts, peering
WHOIS	Registrars, contacts, organization mapping, history
Threat Intel	~40 feeds across 18 categories
Web Graph	10.9 billion hyperlinks for inter-domain analysis
Hosting	Provider identification, infrastructure mapping

The explain_indicator tool also supports CALL explain() for full threat scoring — breaking down which feeds contributed to a score and why. This transparency is critical for HITL workflows where a human needs to understand the reasoning before acting.

Comparison: WhisperGraph vs. Traditional Threat Intel Pipelines

Capability	WhisperGraph MCP	Traditional Stack
Single query interface	✓ Cypher	— 4+ APIs
Cross-entity traversal	✓ Native graph	— Manual JOINs
Programmatic agent signup	✓ 2 HTTP calls	— Browser + CAPTCHA
Built-in investigation templates	✓ 8 prompts	— Custom playbooks
Threat scoring transparency	✓ CALL explain()	— Black-box scores
Self-hosted option	✓ Docker	— Per-tool deployment
Query safety validation	✓ Rule-based	— API-level only

For security teams building agent-powered workflows, the single-interface model eliminates the orchestration layer that would otherwise dominate development time.

Key Takeaways

• 46 billion nodes, one endpoint: DNS, BGP, WHOIS, and threat intelligence unified in a Cypher-queryable graph
• Agent-first signup: Programmatic API key provisioning in 2 HTTP calls — no browser, no CAPTCHA, no waiting
• Six read-only tools: Query, schema exploration, threat assessment, history, and domain variant detection
• Eight investigation prompts: Structured workflows for common security tasks — blast-radius, threat-triage, typosquat-sweep, and more
• Hosted or self-hosted: Use mcp.whisper.security for zero-install, or Docker for on-premises deployments
• Query safety: All Cypher validated against a safety rule set — destructive operations rejected at the protocol level
• Transparent threat scoring: CALL explain() shows exactly which feeds and factors contribute to a threat assessment
• 36B reasons this matters: The graph grows continuously. Your agent always queries the latest internet infrastructure data — no stale snapshots

---

WhisperGraph: whisper.security · GitHub: github.com/whisper-sec/whisper-graph-mcp · Agent Signup: whisper.security/docs/agent-signup · Cypher Reference: whisper.security/docs/cypher-api-reference · Glama: glama.ai/mcp/servers/whisper-sec/whispergraph · Facio MCP docs: facio.bot/docs/mcp