GymCam Analytics

GymCam turns the cameras a gym already has into automatic attendance and trainer-performance analytics — no new hardware, no check-ins.

Feed it the existing CCTV stream and the class schedule; it recognizes trainers, counts attendees per class, and reports what's actually happening: which classes are full, which are dead, and which trainers fill the room.

Why

Gym owners run on gut feeling. Booking software (Mindbody, Glofox) only captures check-ins — and people skip check-ins, so the data is incomplete. Hardware people-counters (Density, V-Count) cost thousands and count bodies without context.

GymCam reuses what's already in the building and maps counts to classes and trainers — the thing that actually drives revenue.

Related MCP server: Iridium MCP Server

What it reports

• Today's summary — classes held, total attendance, top classes

• Trainer attendance — daily/weekly fill rate and no-show rate per trainer

• Class performance — every class ranked by fill rate, so dead classes are obvious

• Revenue insights — most profitable vs. least profitable classes

What you actually get

• Zero install — cameras already do the counting (security cameras are required in nearly every country). No sensors, no mounting, no new hardware.

• No check-in friction — stop making members do a meaningless task; people just show up.

• Class truth — which classes are full and which are dead, not the paper log anyone can fudge.

• Trainer accountability — real fill rate + no-shows per trainer; the "16 becomes 20" rounding dies.

• Occupancy & density — overfull classes and cramped rooms are a pricing / scaling / staffing signal.

• Room optimization — see the big room idle while classes squeeze into the small one; swap and fix.

• Equipment utilization — which machines are actually used; sell, buy, or rearrange.

• Density heatmaps — attraction points and dead zones; change the layout with data.

• Demographics — gender + approximate age breakdown (within GDPR / local law).

• Digital twin — treat the gym as a measurable 3D space; a live model of what's working and what to cut.

• AI-native — an MCP server, so your AI agent reads the data and answers "how's my gym doing today."

Install

One command, straight from this repo (requires uv):

uvx --from git+https://github.com/axelfreeman/gymcamanalytics gymcam

Peek at the tool output without wiring MCP:

uv run scripts/demo.py

Skill (optional)

SKILL.md documents the GymCam tools for any agent that loads skills. Install with a symlink so git pull keeps it fresh:

ln -s "$PWD" ~/.agents/skills/gymcam-analytics

Connect to your AI agent

Same command, different config file per client.

Claude Desktop (claude_desktop_config.json):

{"mcpServers": {"gymcam": {"command": "uvx", "args": ["--from", "git+https://github.com/axelfreeman/gymcamanalytics", "gymcam"]}}}

Codex (~/.codex/config.toml):

[mcp_servers.gymcam]
command = "uvx"
args = ["--from", "git+https://github.com/axelfreeman/gymcamanalytics", "gymcam"]

Cursor (.cursor/mcp.json) and Windsurf (~/.codeium/windsurf/mcp_config.json) use the same JSON block as Claude Desktop.

Claude Code:

claude mcp add gymcam -- uvx --from git+https://github.com/axelfreeman/gymcamanalytics gymcam

API key

Tools require an API key. Get one free at https://gymcamanalytics.com/get-key (100 free lookups, no credit card), then set it:

export GYMCAM_API_KEY=your_key_here

Tools

Status

Pre-launch. The MCP server and tool schema are live; tools return sample data until your gym's cameras are connected. Sign up for access.

License

MIT © 2026 Axel Freeman

Need this done for you?

The stack behind this repo runs as a service: marketing engineering, turnkey — Sprint $900 one-time, Engine $1,900/month, full build $2,900. Scope and prices are published before the first call. If you run a local or multi-site business, the countable version of the same work is here: marketing engineer for local business.