Kerkov

Open-source AI for penetration testing, security research, and CTFs. Kerkov runs in your terminal, executes real tools against real targets, and keeps the engagement — findings, credentials, bulky tool output — recoverable across context compaction and across sessions.

Authorized use only

Kerkov executes commands and performs active security testing. Use it only on systems you own or have explicit written permission to test, and keep every action inside the agreed scope.

This site is being rebuilt

Kerkov's documentation is being rewritten from the source code (issue #200). Only pages verified against the code as it ships today appear in the navigation; everything else is still in the repository but deliberately unpublished. A few sections — worked examples, and the security and contributing guides — are still on the way; where a page isn't here yet, the README fills the gap.

What it is

Kerkov is a conversational CLI that pairs a model of your choosing — Anthropic, Gemini, NVIDIA NIM, OpenRouter, DeepSeek, Ollama or Codex — with a tool suite built for security work: process execution with real PTY sessions, HTTP, browser automation, SBOM and vulnerability scanning, SAST, CVE intelligence, and web search. Every tool call passes through approval rules you control, on an axis independent of how much narration you want.

Where to go next

• Install

  From source, one-line installer, Docker Compose, and the optional extras — get Kerkov running in a few minutes.

• Your first engagement

  A ten-minute walkthrough: launch, scope a target, approve tool calls, and record findings.

• Guides

  Task-shaped how-tos — modes and autonomy, approvals, providers, engagements, skills, memory, and non-interactive use.

• Reference

  Every flag, slash command, and configuration variable, checked against the source.

Install

git clone https://github.com/Kerkov-Security/kerkov.git
cd kerkov
uv sync --locked
uv run kerkov

See Install for the one-line installer, Windows, Docker Compose, and the optional extras.