Keel

Network Diagnostics MCP Server

Probing what lies beneath the surface -- network diagnostics for AI tools.

What It Does

Keel is a Model Context Protocol (MCP) server that gives AI assistants 14 network diagnostic tools. It handles the things you'd normally reach for ping, dig, nmap, or openssl to do -- but exposed as structured, validated MCP tool calls.

Related MCP server: mcp-nettools

Tools

Installation

From PyPI:

pip install keel-mcp

Or isolated with pipx:

pipx install keel-mcp

Usage

Run the server directly (stdio transport):

keel

Claude Code

Register as a local MCP server:

claude mcp add keel -- keel

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "keel": {
      "command": "keel",
      "args": []
    }
  }
}

If installed in a virtual environment, use the full path to the binary:

{
  "mcpServers": {
    "keel": {
      "command": "/path/to/.venv/bin/keel",
      "args": []
    }
  }
}

Security

Keel is designed to be safe for AI-driven use:

• SSRF protection -- http_check resolves hostnames and blocks requests to internal, private, loopback, and link-local IP addresses (including IPv4-mapped IPv6). Cloud metadata endpoints (169.254.x.x) are blocked.

• Input validation -- All inputs pass through validators that reject shell metacharacters, malformed hostnames, and invalid ports before reaching any network call or subprocess.

• Rate limiting -- port_scan enforces a minimum 1-second interval between scans to prevent abuse.

• Subnet restriction -- subnet_scan only allows RFC 1918 private subnets and caps at /20 (4096 addresses) with concurrency limiting.

• No shell injection -- Subprocess calls (traceroute, whois) use exec-style invocation, never shell interpolation.

Development

git clone https://github.com/seayniclabs/keel.git
cd keel
python -m venv .venv
source .venv/bin/activate
pip install -e ".[test]"
python -m pytest tests/ -q

Using with a Gateway

If you're running multiple MCP servers, route them through a gateway like tbxark/mcp-proxy to manage all child processes from a single persistent service. The proxy handles process lifecycle, centralized config, and crash recovery — each server still gets its own SSE endpoint but you manage everything from one config file instead of scattered Claude Code entries.

For a full walkthrough of how this works in practice, see The Hidden Cost of a Loaded MCP Stack on charlieseay.com.

License

MIT