GeekyGramps-MCP

Purpose

A Fastify-based MCP (Model Context Protocol) gateway providing shared, project-scoped memory to multiple AI coding agents (Claude Code, Codex, Antigravity) working against the same homelab. Backed by PostgreSQL, with per-request token enforcement on every route except a health check, and secrets delivered via an Infisical CRD rather than baked into config.

Built so several agents can read/write a common project memory store without stepping on each other, instead of each agent keeping its own disconnected context.

Structure

src/       Fastify server, DB access, project-key validation, tool policy enforcement
scripts/   DB provisioning/init, MCP reachability diagnostics, agent tmux session launcher, request router
test/      Node built-in test coverage for the token gate, tool policy, and project-key validation
deploy/    k3s app-template values, Infisical secret CRD, fail-closed NetworkPolicy

Stack

• Runtime: Node.js, Fastify
• Storage: PostgreSQL
• Protocol: Model Context Protocol (@modelcontextprotocol/sdk)
• Secrets: Infisical (Kubernetes CRD-based sync)
• Deploy target: k3s, NodePort-only exposure, fail-closed NetworkPolicy

Triple-Threat Orchestration System

I got tired of AI tools that just do things. Build something, deploy something, write something — and then I find out three steps later that it made a bad assumption in step one that affects everything downstream.

Triple-Threat is how I push back on that. It’s a structured, gated workflow where three different AI systems each have a specific job, nothing moves forward without passing a gate, and I approve every gate before anything gets built.

The coordinator is Hermes — an AI agent with a built-in learning loop that maintains state, enforces the stage sequence, and keeps everyone on task. I don’t use it as a general assistant. It runs the workflow.

Claude Code does the design and architecture work. Codex handles implementation review and builds the actual runbook. Antigravity (I call it agy) does an independent review of everything Claude Code decided — I make them disagree with each other on purpose, then I decide who’s right.

Any restart, configuration change, DNS update, secret rotation, or infrastructure action still requires my approval.

HexMap — Real-Time Homelab Infrastructure Map 🔥

HexMap is a custom real-time infrastructure visualization system for my FNDHOME homelab. It turns the environment into an interactive operations map, showing hosts, VMs, containers, Kubernetes services, storage, tunnels, monitoring, and service dependencies in one live view.

Unlike traditional monitoring tools, HexMap does not require agents or installs on the systems it tracks. It builds its view from live infrastructure sources, so the map stays current as the environment changes.

Instead of digging through dashboards, manifests, DNS records, and Kubernetes commands, HexMap gives me a visual control plane for understanding what exists, where it runs, what it depends on, and what could break if something goes down.

The design is built around operational clarity: services are grouped by infrastructure layer, dependency lines show relationships between systems, and health indicators surface what needs attention.

The goal is not just to display the homelab — it is to make the stack understandable, navigable, and eventually actionable.

HexMap is the kind of tool I wanted as an admin: a living infrastructure map that can grow from documentation into an ops console.

FNDHOME Infrastructure Architecture Map

High-level architecture diagram of my FNDHOME infrastructure, showing the relationship between internet ingress, firewall/security services, Proxmox hosts, virtual machines, LXCs, storage, databases, backups, monitoring, and public access tunnels. This diagram documents the core design of the environment, including separation of critical infrastructure from application workloads, external database placement, backup paths, SSO routing, and tunnel-based public access.

Kubernetes Workload Inventory

A live Kubernetes workload view showing application pods running inside the homelab namespace. This snapshot represents the migrated self-hosted application stack, including services such as Authentik, Immich, Nextcloud, Jellyfin, Gitea, Vaultwarden, n8n, OnlyOffice, WordPress, Syncthing, and other internal tools. It highlights the move from traditional Docker-hosted apps into a more centralized Kubernetes-based operating model.