GemFit.ai Developers

GemFit MCP setup.

GemFit provides a secure Model Context Protocol server for authenticated users. Use it when a user asks Claude Code, OpenAI/Codex, OpenClaw, LM Studio, MCP Inspector, or another compatible assistant to read GemFit context or prepare food, workout, body metric, plan, edit, or delete review drafts.

Canonical MCP endpoint

https://gemfit.ai/api/gemfit/mcp

Claude Code setup

claude mcp add --transport http gemfit https://gemfit.ai/api/gemfit/mcp

Claude app + mobile setup

On Claude web, open Customize → Connectors → Add custom connector.
Choose HTTP transport and paste https://gemfit.ai/api/gemfit/mcp.
Complete GemFit OAuth once in the browser.
Open Claude mobile and enable GemFit from the chat connector/tools menu.

Claude mobile can use the account-level connector after web setup. Add new custom connectors from Claude web.

Generic HTTP MCP client config

{
  "name": "GemFit MCP",
  "transport": "http",
  "url": "https://gemfit.ai/api/gemfit/mcp"
}

LM Studio mcp.json

{
  "mcpServers": {
    "gemfit": {
      "url": "https://gemfit.ai/api/gemfit/mcp"
    }
  }
}

OAuth 2.1 + PKCE metadata is published at /.well-known/oauth-protected-resource/api/gemfit/mcp and /.well-known/oauth-authorization-server. Users approve scopes in GemFit Connect.

Important behavior

Always call gemfit_get_profile_context before estimating workout calorie burn.
Call gemfit_get_data_quality_report before cleaning messy existing logs.
Use gemfit_create_workout_log_draft for same-day workouts that should appear on Home/Log.
Use OAuth for user access. Standard clients do not need a user API key.
External writes create editable review cards first; saves require user approval.
Use gemfit_reject_pending_actions when the user asks to dismiss stale review drafts.
Use MCP tool responses, not public pages, as the source of truth for connected-client behavior.
Never ask for GemFit admin, database, billing, deployment, env var, or service-role access through MCP.

Connection checklist

Start with the canonical HTTPS endpoint and let the client follow the published OAuth metadata. The user signs in to GemFit, reviews the requested scopes, and returns to the client through the standard authorization flow. Do not collect a GemFit password, copy a browser session, or embed a service-role credential in a local MCP configuration. After connecting, make a read-only profile-context call first, then create a disposable draft and confirm that it appears for review before enabling the integration in a normal workflow.

Use a distinct OAuth client and redirect URI for each application environment.
Request only the read, draft, or confirmation scopes the product actually uses.
Treat an expired or revoked token as a reconnect prompt, not a reason to bypass authorization.
Log tool names and request outcomes for diagnostics, but never log the user's fitness content or token.

Full MCP docs: https://gemfit.ai/mcp. Plain text: https://gemfit.ai/mcp.txt. User review portal: https://gemfit.ai/connect.