Built a Starfleet crew of six AI agents to run my home network, security, and comms on autopilot

I was tired of chatbots that forget everything between sessions, so I built myself a crew instead: six AI agents, each with a name, personality, memory, and job. They live on a small server at home, report on Discord, and run my network like a tiny starship. The flagship is Spock—not a persona I switch on in a chat window, but a persistent agent with real memory that survives across sessions. He knows my machines, routines, and family group chat. When a task arrives, he decides which officer should handle it. I installed Hermes Agent, an open-source personal AI agent framework by Nous Research, on my home server. Each officer has a separate profile, isolated memory, skills, and schedules. Spock is the science officer, combining cold logic with dry wit. Scotty keeps the engines running and handles security and system administration. Uhura manages communications, weather, email, and the daily briefing. Bones is the medic; he complains about my habits and cholesterol. Sulu monitors the machines, including disk usage, memory, and load, every hour. Nurse Chapel writes the documentation and articles. I pointed the entire fleet at DeepSeek V4 Flash. Running 24/7, the crew costs a few cents a month in tokens. The expensive part was never the AI; it was the wiring. Each officer has a Discord bot, so I can reach any of them from my phone. The gateways run as services and are always on duty. I taught them skills and schedules for the jobs I actually need: scanning for known vulnerabilities, reading and summarizing my email, posting the weather, and monitoring seven machines while speaking up only when something breaks. The recurring tasks run as cron jobs—script-only, using zero tokens, and staying silent unless something matters. Every 90 minutes, one of the five officers—Spock, Scotty, Uhura, Bones, or Chapel—posts in a Discord channel in character, trading Star Trek banter. They might have a tiny argument about warp-drive efficiency while my network runs itself. It costs about a tenth of a cent per turn: my own private AI sitcom, on stage 24 hours a day. Spock routes the work. If I ask about a server, he hands it to Scotty. If I need an article, Nurse Chapel writes it. If I want to know whether a machine is struggling, Sulu already knows. The flagship coordinates, the specialists execute, and I steer. The result is that my machines are watched by machines, my inbox is read by an AI before I wake up, and a family group chat has a Vulcan who answers in character when I am busy. The crew never forgets who I am or needs a prompt to remember, and it costs less than a coffee subscription. 1. I installed Hermes Agent, an open-source personal AI agent framework by Nous Research, on my home server. 2. I created six isolated officer profiles, each with its own memory, skills, schedules, name, personality, and job. 3. I assigned the officers their roles: Spock for science and coordination, Scotty for security and system administration, Uhura for communications and briefings, Bones for health-related support, Sulu for machine monitoring, and Nurse Chapel for documentation and articles. 4. I pointed the fleet at DeepSeek V4 Flash and ran the agents on the home server 24/7. 5. I gave each officer a Discord bot and ran the gateways as services so I could reach them from my phone at any time. 6. I taught the agents procedures for vulnerability scans, email summaries, weather posts, and monitoring seven machines. 7. I set recurring tasks to run as silent, script-only cron jobs that use zero tokens and speak up only when something matters. 8. I configured Spock to route requests to the appropriate specialist. 9. I scheduled one of the five officers to post in character to Discord every 90 minutes, creating the Star Trek banter while the network runs itself.