VaultCat

Full-lifecycle HashiCorp Vault penetration testing toolkit — recon, hijack, escalate, exfiltrate, pivot, and persist across three operational modes.

• External Recon — zero-knowledge Vault fingerprinting, TLS/CORS/header audit, version→CVE matching, auth surface mapping
• Authenticated Assessment — token capability audit, policy analysis, KV enumeration, TTL governance, auth config audit
• Local Post-Exploitation — filesystem + git credential scanning (56 patterns), cross-file correlation, AppRole/DB validation
• Active Execution — 30 state-changing modules (privilege escalation, secret exfiltration, persistence, pivot, CVE exploitation)
• AI Agent — ReAct-loop autonomous pentest agent with multi-provider LLM, attack tree walker, and mutation engine
• MCP Server — 52 tools exposed via FastMCP for Claude Desktop and other AI clients

Ethics: Use this tool only on systems you own or have explicit permission to assess. Read-only modules run freely; state-changing and destructive operations require --confirm-active. No brute-force or password cracking.

Install

# Install from GitHub (PyPI package coming soon)
pip install git+https://github.com/muhammedkurtoglu0/vaultcat.git

# Or with uv (isolated environment, PATH-linked)
uv tool install git+https://github.com/muhammedkurtoglu0/vaultcat.git

# Dev install (clone + venv):
git clone https://github.com/muhammedkurtoglu0/vaultcat.git
cd vaultcat
uv sync
uv run vaultcat chat

Quick Start

# Unauthenticated recon
vaultcat scan --target https://vault.example.com:8200

# Authenticated audit
vaultcat scan --target https://vault.example.com:8200 --token hvs.xxx --capability-audit

# Local credential hijacking
vaultcat hijack ./my-repo --validate-token --target https://vault.example.com:8200

# AI-powered pentest chat (terminal)
vaultcat chat

# AI chat with desktop GUI
vaultcat chat --ui desktop

# MCP server (for Claude Desktop integration)
vaultcat mcp

Documentation

Connect to Claude Desktop

vaultcat mcp --transport stdio

Then add to your Claude Desktop config:

{
  "mcpServers": {
    "vaultcat": {
      "command": "vaultcat",
      "args": ["mcp", "--transport", "stdio"]
    }
  }
}

52 pentest tools appear in Claude's toolbox. Full MCP guide →

Supported LLM Providers

Project Layout

main.py                 CLI entry point (Typer, 5 commands)
vault_cli.py            Typer app definition
core/                   Reporting, risk scoring, TLS config
reconnaissance/         14 unauthenticated recon scanners
scanners/               10 authenticated assessment scanners
credential_hijacking/   7 file/git scanning + correlation modules
active_execution/       30 state-changing assessment modules
  modules/
    cloud/              AWS/Azure/GCP exploitation
    database/           DB credential harvest + exploitation
    general/            CVE scanner, agent sidecar, DoS
    persistence/        Backdoors + audit manipulation
    pivot/              Cross-service lateral movement
    seal/               Seal/unseal + key exfiltration
    secrets/            KV dump, PKI, Transit, Raft storage
    token/              Priv esc, token/policy exploits, K8s/JWT/AppRole
ai_core/                LLM agent, MCP server, chat UI, planning, session
vaultcat-lab/      Docker-based lab (Vault 1.15.3 + PostgreSQL 16)
tests/                  23 test files, 600+ tests

Vault Pentest Lab

cd vaultcat-lab
docker compose up -d
./scripts/setup-lab.sh
source lab-tokens.env

Full lab guide →

License

MIT — see LICENSE.