Brian Christner's Human OS

Table of content

Brian Christner is a Docker Captain, cloud architect, and Chief Online Gaming at Jackpots.ch in Switzerland. He co-founded 56K.Cloud and created one of the most-starred Docker monitoring stacks on GitHub (4,000+ stars). When not containerizing everything, he’s mountain biking in the Swiss Alps and obsessively tracking his performance data.

Christner built Human OS to solve a specific problem: professionals toggle between apps 1,200 times per day. That’s four hours weekly just flipping between tabs. Human OS uses AI and MCP servers to create one interface for all your tools.

Background

Docker Captain since the early days of the program
Cloud architect specializing in DevOps, monitoring, and containers
Created the docker-compose Prometheus monitoring stack
Speaker at DockerCon, Startup Nights, and container conferences
Based in Zurich, originally from Arizona

GitHub | Twitter | Blog

Human OS Architecture

Human OS operates on four principles:

Principle	Implementation
Centralize context	Single conversational interface
Automate orchestration	AI coordinates across apps via MCP
Personalize execution	System learns user patterns
Free mental attention	No more app-switching

The architecture:

[You] → [Claude Desktop] → [MCP Servers] → [External Services]
              ↑                    ↓
         AI (brain)           Spine (integration)

AI serves as the brain, interpreting requests and planning steps. MCP functions as the spine, connecting tools through standard interfaces. Together they eliminate manual tab-switching and context loss.

MCP Server Stacking

The power of Human OS comes from combining multiple MCP servers. Christner demonstrates this with fitness tracking.

Individual strengths:

Server	Data Type
Garmin MCP	Sleep, HRV, stress, recovery
Strava MCP	Segment times, routes, competition

Combined intelligence unlocks:

Recovery vs. performance correlation
Hidden training pattern detection
Sleep-to-performance linkage
Predictive readiness (“Am I ready for a PR attempt today?”)

Measured Results

Metric	Single MCP	Dual MCP
PR attempt success	35%	71%
Injury prevention	60%	85%
Recovery timing accuracy	80%	95%

The insight: stacking MCPs creates compounding value. One server gives you data. Two servers give you intelligence.

Setup

Christner runs Human OS in Claude Desktop with two MCP servers:

{
  "mcpServers": {
    "garmin": {
      "command": "uvx",
      "args": ["garmin_mcp"],
      "env": {
        "GARMIN_EMAIL": "your@email.com",
        "GARMIN_PASSWORD": "your-password"
      }
    },
    "strava": {
      "command": "node",
      "args": ["/path/to/strava-mcp/index.js"],
      "env": {
        "STRAVA_CLIENT_ID": "your-client-id",
        "STRAVA_CLIENT_SECRET": "your-secret"
      }
    }
  }
}

Garmin MCP: github.com/Taxuspt/garmin_mcp (Python 3.12 via uvx)

Strava MCP: github.com/r-huijts/strava-mcp (Node.js)

Business Application

The same pattern applies to work. Christner describes a professional workflow:

CRM MCP + Calendar MCP + Slack MCP

Before Human OS: Open Salesforce, search client, copy notes. Open calendar, check history. Open Slack, find thread. Switch between windows to compile context. 15 minutes of prep before a call.

After Human OS: Single prompt to Claude: “Prepare me for my call with Acme Corp in 10 minutes.”

The AI reads CRM data, pulls calendar history, finds relevant Slack threads, and presents a unified brief. Context switching drops to zero.

Key Takeaways

Principle	Implementation
One interface	Claude Desktop as unified entry point
Stack MCPs	Combine servers for compounding intelligence
Measure results	Track before/after metrics
Start with one domain	Pick fitness or work, not both

Getting Started

Week 1: Single MCP

Install Claude Desktop
Pick one MCP server (Garmin, Strava, or a work tool)
Configure the connection
Ask questions about your data

Week 2: Add second MCP

Install a complementary server
Ask questions that span both data sources
Note what insights emerge from the combination

Week 3: Expand to work

Add CRM, calendar, or communication MCPs
Build a “meeting prep” workflow
Measure time saved