Arbox MCP
An MCP server that connects Claude AI to the Arbox gym management platform. 43 endpoints covering members, scheduling, leads, and billing.
The Problem
Gym managers spend hours on repetitive tasks — looking up member info, checking schedules, managing leads. The Arbox platform has an API, but no AI integration existed.
What I Built
An MCP (Model Context Protocol) server that lets Claude AI interact with Arbox directly. It exposes 43 endpoints covering:
- Member management — search, create, update, view memberships
- Scheduling — view classes, book sessions, manage waitlists
- Lead tracking — create leads, update status, add notes
- Reporting — pull attendance, revenue, and membership reports
The server handles Hebrew-first data, date formatting, and Arbox’s API quirks (case-sensitive params, type coercion, silent failures) transparently.
Technical Approach
Built as a standard MCP server that any MCP-compatible AI client can use. The main challenge was mapping Arbox’s inconsistent API responses into clean, typed tool definitions that make sense to an AI agent.
What I’d Do Differently
Would add response caching for frequently-accessed data (schedules, membership types) to reduce API calls and improve response times.