Connectivity Diagnostics MCP Server

An MCP server that gives your AI assistant real network diagnostic capabilities. Instead of guessing about your network, the AI can actually run tools like ping, traceroute, DNS lookups, and TLS certificate checks — right on your machine — and interpret the results for you.

Works on Windows, macOS, and Linux.

Important: This server is designed to run locally on your machine. The diagnostic tools (ping, traceroute, etc.) execute against your local network stack — that's the whole point. If you ran the server on a remote host, the tools would diagnose that host's network, not yours. For this reason, you need an MCP client that supports local stdio servers (see compatibility notes below).

---

Try These First

Once installed, try asking your AI assistant questions like these:

• "How is my connection to google.com? Run every diagnostic you have."
• "Why might my connection to AWS us-east-1 be slow? Investigate."
• "Check the SSL certificate on mydomain.com — is it about to expire?"
• "Run a full self-diagnostic of my network configuration."
• "What DNS server am I using, and is it resolving correctly?"
• "Trace the route to cloudflare.com and identify any high-latency hops."
• "Break down the HTTP timing for api.example.com — where is the delay?"
• "What devices are on my local network right now?"

---

Available Tools

Tool	What It Does
ping	ICMP ping — latency and packet loss to any host
curl	HTTP headers — check reachability and status codes
tracert	Route tracing — hop-by-hop path to a host
nslookup	DNS lookup — verify domain name resolution
ipconfig	Local adapter config — IPs, gateway, DNS servers
routing_table	System routing table — how packets are directed
http_timing	Detailed HTTP timing: DNS, TCP, TLS, TTFB breakdown
ssl_check	TLS certificate details and validity
arp_table	LAN device discovery via ARP cache
mtr	Combined ping + traceroute with per-hop loss stats

---

Prerequisites

Install the MCP Python package:

pip install "mcp[cli]"

Download connectivity_mcp_server.py and note its full path. You will need it in the configuration steps below.

---

Setup: Claude Desktop

Claude Desktop connects to local MCP servers via stdio. You add the server to a JSON config file and restart the app.

1. Find the config file

OS	Path
Windows	%APPDATA%\Claude\claude_desktop_config.json
macOS	~/Library/Application Support/Claude/claude_desktop_config.json

Paste the path into Explorer or Finder to open the folder. If the file does not exist, create it.

2. Add the server

Open claude_desktop_config.json and add or merge the following. Replace the path with the actual location of connectivity_mcp_server.py on your machine:

Windows:

{
  "mcpServers": {
    "connectivity-diagnostics": {
      "command": "python",
      "args": [
        "C:\\path\\to\\connectivity_mcp_server.py"
      ]
    }
  }
}

macOS / Linux:

{
  "mcpServers": {
    "connectivity-diagnostics": {
      "command": "python3",
      "args": [
        "/path/to/connectivity_mcp_server.py"
      ]
    }
  }
}

If you already have other MCP servers configured, add the "connectivity-diagnostics" block inside the existing "mcpServers" object — don't replace the whole file.

3. Restart Claude Desktop

Fully quit and relaunch the app. The MCP server starts automatically when Claude needs it.

4. Verify

Ask Claude: "What network diagnostic tools do you have available?"

---

Setup: Cursor

Cursor supports local MCP servers through a mcp.json config file.

1. Create the config file

You can configure MCP servers globally or per-project:

Scope	Path
Global	~/.cursor/mcp.json
Project	.cursor/mcp.json (in project root)

2. Add the server

Create or edit the mcp.json file. Replace the path with the actual location of connectivity_mcp_server.py on your machine:

Windows:

{
  "mcpServers": {
    "connectivity-diagnostics": {
      "command": "python",
      "args": [
        "C:\\path\\to\\connectivity_mcp_server.py"
      ]
    }
  }
}

macOS / Linux:

{
  "mcpServers": {
    "connectivity-diagnostics": {
      "command": "python3",
      "args": [
        "/path/to/connectivity_mcp_server.py"
      ]
    }
  }
}

3. Restart Cursor

Relaunch the editor or reload the window. You can verify the server is connected under Settings > Tools & Integrations > MCP Tools.

---

What About ChatGPT?

ChatGPT's MCP support (Developer Mode) only connects to remote servers over HTTPS — it does not support local stdio. Since this server is designed to diagnose your machine's network, running it remotely would defeat the purpose. ChatGPT is not compatible with this server at this time.

---

Troubleshooting

The AI doesn't see the tools after restart:

• Double-check the file path in your config — use \\ on Windows.
• Confirm python (or python3) is on your system PATH: run python --version in a terminal.
• Check the client's MCP logs if available.

tracert or mtr times out:

• This is normal for hosts that block ICMP. The timeouts are generous (45s for tracert, 60s for mtr).
• Try a well-known host like 8.8.8.8 to confirm the tool works.

mtr command not found:

• Windows: Install WinMTR and ensure it is on your PATH.
• macOS: brew install mtr
• Linux: sudo apt install mtr (Debian/Ubuntu) or sudo dnf install mtr (Fedora/RHEL)