GemFit.ai Docs

AI fitness logging with user-approved tools.

GemFit combines mobile fitness logging with a secure MCP server. Approved AI clients can use the MCP to read the authenticated user's profile, goals, day summaries, recent trends, and logs, then create editable review drafts for messy meals, workouts, body metrics, plans, updates, and deletes.

Find the MCP

Remote MCP: https://gemfit.ai/api/gemfit/mcp
OAuth:      https://gemfit.ai/.well-known/oauth-protected-resource/api/gemfit/mcp
Docs:       https://gemfit.ai/mcp
Plain text: https://gemfit.ai/mcp.txt
Connect:    https://gemfit.ai/connect

Claude Code command:

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

LM Studio mcp.json:

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

Logging flow

User asks an AI client to log food or workouts.
Client calls gemfit_get_profile_context and gemfit_get_day_summary.
Client calls gemfit_get_data_quality_report before cleaning or repairing existing logs.
Client calls draft tools such as gemfit_create_food_log_draft or gemfit_create_workout_log_draft.
Same-day workouts should use workout log drafts, not future plan drafts.
GemFit shows editable review cards.
User approves in GemFit Connect, or explicitly approves a specific draft in chat when logs.confirm is granted.
Stale cards can be rejected with gemfit_reject_pending_actions after explicit user rejection.
Standard clients use OAuth. GemFit does not require users to paste personal API keys into local configs.

Security boundaries

GemFit MCP exposes user-owned fitness data only. It cannot access Systalog admin, billing admin, deployments, Supabase service role, schema changes, env vars, role changes, email systems, or other users.