kali-mcp
# kali-mcp
An MCP server that gives AI assistants access to a full Kali Linux environment running in Docker. Connect it to Claude Desktop, Claude Code, or any MCP-compatible client and let the AI run security tools like nmap, sqlmap, hydra, nikto, gobuster, and more.
## How it works
```
Claude <--stdio--> MCP Server <--Docker API--> Kali Linux Container
```
The MCP server manages a Docker container running Kali Linux. It exposes tools that let the AI start/stop the container, execute commands, and transfer files — all through the [Model Context Protocol](https://modelcontextprotocol.io).
## Tools
| Tool | Description |
|---|---|
| `container_start` | Start the Kali Linux container |
| `container_stop` | Stop and remove the container |
| `container_status` | Check if the container is running |
| `execute_command` | Run any shell command inside Kali |
| `upload_file` | Write a file to the container |
| `download_file` | Read a file from the container |
| `list_files` | List directory contents |
## Pre-installed tools
nmap, nikto, gobuster, sqlmap, hydra, john, hashcat, metasploit-framework, dirb, enum4linux, curl, wget, python3, wordlists, and more.
## Prerequisites
- [Node.js](https://nodejs.org/) v18+
- [Docker Desktop](https://www.docker.com/products/docker-desktop/) (must be running)
## Setup
### 1. Clone and install
```bash
git clone https://github.com/Hannes221/kali-mcp.git
cd kali-mcp
npm install
npm run build
```
### 2. Build the Kali Docker image
```bash
cd docker
docker compose build
cd ..
```
This downloads and builds the Kali Linux image (~4GB). It takes a few minutes on the first run.
### 3. Connect to your AI client
#### Claude Desktop
Open **Settings** > **Developer** > **Edit Config** and add the `kali` server to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"kali": {
"command": "node",
"args": ["/absolute/path/to/kali-mcp/dist/index.js"]
}
}
}
```
Replace `/absolute/path/to/kali-mcp` with the actual path where you cloned the repo.
Then **restart Claude Desktop**.
#### Claude Code
If you cloned the repo, the `.mcp.json` file is already included. Just open the project directory in Claude Code:
```bash
cd kali-mcp
claude
```
Or add it manually to your Claude Code config:
```bash
claude mcp add kali node /absolute/path/to/kali-mcp/dist/index.js
```
### 4. Verify
In a conversation, ask the AI to:
> Start the Kali container and run `nmap --version`
It should call `container_start`, then `execute_command` with `nmap --version` and return the version output.
## Usage examples
Once connected, you can ask the AI things like:
- "Scan 192.168.1.0/24 for open ports"
- "Run nikto against http://target.com"
- "Use sqlmap to test http://target.com/page?id=1 for SQL injection"
- "Crack these hashes with john"
- "Enumerate SMB shares on 10.0.0.5"
The AI will use the appropriate Kali tools and interpret the results for you.
## Development
```bash
npm run dev # Run with tsx (auto-compiles TypeScript)
npm run build # Compile to dist/
npm start # Run compiled version
```
## Customizing the Kali image
Edit `docker/Dockerfile` to add or remove tools, then rebuild:
```bash
cd docker
docker compose build
```
## Security
- The container runs with default Docker isolation (no `--privileged` flag)
- Network mode is `bridge` by default
- The container is ephemeral — no persistent volumes
- **Only use this for authorized security testing, CTF challenges, and educational purposes**
## License
MIT
TDQS
Scored across 7 tools
Each tool has a clearly distinct purpose with no overlap: container management (start/status/stop), file operations (download/list/upload), and command execution. The descriptions explicitly differentiate their functions, making misselection unlikely.
All tools follow a consistent verb_noun pattern with snake_case naming (e.g., container_start, execute_command, list_files). The naming convention is uniform across all seven tools, making them predictable and easy to understand.
Seven tools is well-scoped for managing a Kali Linux Docker container, covering essential operations like container lifecycle, file management, and command execution. Each tool earns its place without redundancy or bloat.
The toolset provides complete coverage for the domain: container lifecycle (start/status/stop), file operations (upload/download/list), and command execution. There are no obvious gaps, enabling agents to perform all necessary tasks for security testing workflows.