Skip to main content
The FITsociety Model Context Protocol (MCP) server lets approved AI clients connect to FITsociety data through a controlled tool interface. AI assistants can use the MCP server for two subject surfaces:
  • Company/coach access: a Manager or Admin coach approves tools for company-scoped workflows, such as client search, bookings, invoices, notes, nutrition plans, nutrition logs, and selected write actions.
  • Client AI access: a client approves tools that can only act on that client’s own records in the active company context. The client must be connected to that company, have client AI access enabled, and the company MCP integration feature must be enabled.

​
MCP endpoint

MCP clients call this endpoint with JSON-RPC 2.0 payloads over stateless HTTP POST requests.

​
Versions and tool results

The server reports its release in serverInfo.version. This release number is separate from the negotiated MCP protocol version and the /mcp/v1 endpoint. Compatible additions increment the minor version; compatible fixes increment the patch version. Breaking tool-contract changes require a major version. Starting with server release 1.0.0, every tool returned by tools/list includes an outputSchema for its successful structuredContent. Schemas describe the allowed fields, nested objects, and arrays. Fields can be omitted when empty or not applicable. The current schemas leave leaf value types unconstrained. Check isError before interpreting a result as a success.

​
What it is for

Use FITsociety MCP when an AI assistant needs permissioned access to FITsociety tools, for example:
  • Search for clients
  • Summarize bookings and attendance
  • Inspect invoices and subscriptions
  • Read health-related data when explicit consent is granted
  • Create or update allowed records through explicitly enabled write tools
  • Let clients review or update their own records through client AI access

​
Security model

Every connection is scoped to one grant. Company/coach grants are scoped to one company and owner coach and require the company MCP integration feature. Client AI grants are scoped to one company and one client and require the company MCP integration feature, the client AI access toggle for that company, and the company client MCP policy. The grant controls which modules are available, whether the connection is read-only, which write tools are enabled, and whether health or private communication data is allowed.