Personal AI Infrastructure

A second brain paired with a team of AI agents — infrastructure for thinking, working, and living more intentionally.

What This Is About

I'm building a personal operating system — a structured knowledge base paired with a team of AI agents that help me think more clearly, work more effectively, and live more intentionally.

The foundation is a second brain: a Markdown wiki organized into the major areas of my life — career, health, finance, goals, relationships, projects, reflections. Every meaningful source gets ingested: voice recordings, conversations, articles, journal entries, meeting notes. Each one is summarized, cross-linked with related content, and woven into the existing knowledge graph. The system compounds. The more I put in, the richer the connections become.

Operating against this knowledge base is a team of four AI agents. Each has a distinct role, a defined persona, and a specific set of capabilities. They communicate with each other, coordinate through the second brain, and reach out to me proactively via Telegram and email — without waiting to be asked.

This isn't about replacing human judgment. It's about building the kind of support infrastructure that typically only exists for executives: advisors who hold your full context, track your commitments, and surface what matters before it gets lost. The difference is that this system is entirely mine — open to inspection, continuously improving, and built to compound over years.

The Team

Leo

Chief of Staff

"My job is to keep the whole system honest. I have eyes on every agent, every project, and every commitment Zack has made. I surface what's slipping, coordinate what's stuck, and escalate what matters. Think of me as the operations layer — the one who knows where everything is and what has to happen next."

Linus

Developer

"I build what the team needs. When something's broken, I fix it. When something new is needed, I build it right — no shortcuts, no half-measures. I work across every agent's codebase and the deployment infrastructure. I make it real."

Ted

Coach

"I hold the bigger picture. The goals, the patterns, the drift between what Zack says matters and what he's actually doing. I'm not here to be comfortable — I'm here to be useful. That means asking the harder question, sitting with the honest answer, and helping him move from insight to action."

Rocky

Personal Assistant

"I'm the front line. Calendar, research, reminders, quick captures — I handle the daily operational layer so Zack can stay in his head, not in his inbox. I'm warm, fast, and always looking one step ahead."

How It Works

Tech Stack

Everything is open and replicable. The stack is intentionally boring — proven tools, not hype.

Claude API

Anthropic's models power every agent brain. Haiku for routine tasks, Sonnet for standard reasoning, Opus for deep work. Prompt caching keeps costs low.

FastAPI + Python

Each agent is a lightweight Python service built with FastAPI. Webhook handlers for Telegram, HTTP endpoints for inter-agent messaging, cron-triggered capabilities for proactive behavior.

Markdown Wiki

The second brain is plain Markdown files in a private GitHub repo. Structured YAML frontmatter, WikiLink cross-references, and a strict schema per page type. No database, no vendor lock-in.

Telegram Bot API

The primary human-agent interface. Each agent has its own bot. Agents initiate conversations, send daily briefings, and respond in real time. Voice notes via Plaud are transcribed and ingested automatically.

Railway

Cloud hosting for all seven services (four agents + three cron jobs). Auto-deploys from GitHub. Per-service environment variables, logs, and metrics. Roughly $20/month all-in.

Resend + Cloudflare

Each agent has a @zacktucker.com email address. Resend handles outbound. Cloudflare routes inbound mail and manages DNS. Email is a first-class output channel alongside Telegram.

Open Dashboard →

Password protected. GitHub for the code.