Skip to content

Back to the catalog

i8void/cybersleuth
github.comrepo

i8void/cybersleuth

CyberSleuth is an open-source MCP server that exposes 34 OSINT/cyber-investigation capabilities to Claude Desktop or Claude Code via natural language, covering infrastructure analysis (DNS, WHOIS, reverse DNS, AS/hosting intel), certificate transparency (CertSpotter/Censys), threat intel correlation (Shodan, VirusTotal, URLScan.io), tech-stack and ATS/GitHub org recon, and privacy/jurisdiction compliance analysis (GDPR/CCPA/LGPD/PIPL/PIPEDA). It also includes an LLM/AI surface reconnaissance module that passively fingerprints LLM-powered apps and runs authorization-gated OWASP LLM Top 10 probes. All external API keys are optional, with graceful degradation when unset.​

TL;DR An MCP server giving Claude 34 OSINT tools for domain, company, and LLM-surface reconnaissance across DNS, certs, Shodan, VirusTotal, and privacy/jurisdiction analysis.
Document

What it is

CyberSleuth (https://github.com/Mar8x/cybersleuth) is a locally-run MCP server exposing 34 OSINT tools (and 9 MCP resources) to Claude Desktop/Code over stdio. Beyond the capability list the summary gives, the page provides the concrete install path, the ten environment-variable-gated API integrations, and the operational-security constraints an engineer needs before pointing it at real targets.

What to do

Install with uv (recommended) or pip:

git clone https://github.com/Mar8x/cybersleuth.git
cd cybersleuth
uv sync

Register with Claude Code:

claude mcp add cybersleuth -- uv run --directory /absolute/path/to/cybersleuth cybersleuth

Or add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) with a mcpServers.cybersleuth block invoking uv run --directory ... cybersleuth, passing keys via env.

All keys are optional; a tool missing its key returns a structured error with a signup link. The ten recognized keys: SHODAN, URLSCAN, VIRUSTOTAL, BUILTWITH, BRAVE, TAVILY, PERPLEXITY, CERTSPOTTER, CENSYS, GITHUB. To drive the agent, load cybersleuth.md as the system prompt / project instructions (it holds the persona, methodology, example queries) — it is also served as an MCP prompt. Standalone methodology guides live in docs/ (ethics.md, methodology.md, per-target workflow files, dns-twist.md, threat-surface.md), several also exposed as cybersleuth:// resources.

Preconditions / Affected

Python 3.10+, plus uv or pip. Passive DNS / reverse-IP / reverse-CNAME runs through ip.thc.org with no key but is rate-limited to ~0.5 req/s — the practical throughput ceiling for bulk subdomain work. Ransomware victim listings come from ransomware.live via manual web fetch, not an API.

Caveats

The page's own operational-security notes matter for authorized use:

  • All external API queries may be logged by the respective services, which track your IP and usage patterns — consider approved proxies for sensitive research.
  • No persistent storage of investigation results (nothing is cached locally).
  • Keys are read from environment variables only, never hardcoded.
  • Identity is ambiguous: the catalog title reads i8void/cybersleuth but every clone URL on the page points to github.com/Mar8x/cybersleuth. Verify the repo owner before trusting the source. License is MIT.

Organisations named here

  • CCPA
  • GDPR
  • RFC 9116

Technologies named here

  • CyberSleuth
  • Shodan
  • VirusTotal
  • URLScan.io
  • BuiltWith
  • CertSpotter
  • Censys
  • Brave Search
  • Tavily
  • Perplexity
  • ip.thc.org
  • Ransomware.live
  • GitHub
  • IPinfo
  • Public WHOIS servers
  • Public DNS services
  • Model Context Protocol
  • Claude Desktop
  • Claude Code
  • Python
  • OWASP Top 10 for LLM Applications

Techniques named here

  • Favicon hash generation
  • DNS enumeration
  • WHOIS investigation
  • Reverse DNS
  • AS intelligence with hosting/cloud detection
  • Passive DNS via ip.thc.org
  • security.txt lookup
  • Multi-source threat intelligence correlation
  • ATS discovery
  • Job-posting keyword extraction
  • GitHub org recon
  • Wayback Machine fallback
  • Passive fingerprinting of LLM-powered apps
  • Benign chat probe
  • Authorization-gated OWASP LLM Top 10:2025 probes
  • Privacy policy discovery and OPP-115 classification
  • Jurisdiction detection with compliance gap analysis
  • AI training clause detection
  • Tracker scan with enforcement-backed contradiction detection
  • People investigation methodology
  • CV claim verification
  • Content character profiling
  • DNS-twist look-alike/permutation methodology

Related in OSINT