Private Developer Network

A Headscale-controlled private network that links Windows and macOS workstations with application servers, K3s infrastructure, reverse proxies, and a local AI inference server.

HeadscaleTailscaleCaddyUbuntumacOSWindows

Abstract private network topology connecting workstations to K3s and AI servers

Context

The Problem

Development machines, application infrastructure, and local AI services needed private connectivity across different operating systems without exposing each service directly to the public internet.

System flow

Architecture

Windows Workstation

macOS Workstation

Headscale Control Server

Tailscale Network

K3s Server

AI Server

Applications

Local LLMs

Implementation

Engineering

Set up a self-hosted Headscale control server for a Tailscale-compatible private network.
Used authentication keys to enroll Windows and macOS machines, application servers, and K3s infrastructure.
Enabled private access to local AI inference and application services.
Connected Caddy reverse-proxy routes across the private network.
Investigated routing and connectivity across heterogeneous machines and services.

What required care

Challenges

Troubleshooting routing and reachability between workstations, servers, and Kubernetes workloads.
Coordinating device enrollment and connectivity across Windows, macOS, and Ubuntu.
Keeping reverse-proxy access aligned with private-network routes.

Technology

Stack

Status: Personal Project

HeadscaleTailscaleCaddyUbuntumacOSWindows