TrainingDojo MCP for ChatGPT: Approval Status and OpenAI API Setup

After OpenAI approves and publishes the TrainingDojo listing, ChatGPT can use it through the app directory on the web or desktop. The normal setup is a Connect button followed by TrainingDojo sign-in and approval; developers can still use the Responses API now.

Before You Connect

• An active TrainingDojo Pro subscription.
• At least one platform connected in Settings.
• The TrainingDojo MCP endpoint: https://trainingdojo.app/api/mcp.

Your TrainingPeaks, Intervals.icu, and Strava credentials stay encrypted in TrainingDojo and are never sent to the AI app. The normal browser flow does not require you to create or handle a developer key. The AI app and provider you choose handle all conversation inference and generation under your provider account. TrainingDojo MCP never calls TrainingDojo's AI models or uses TrainingDojo AI credits. TrainingDojo receives authenticated tool inputs and returns data or action results, not the rest of your provider conversation. TrainingDojo's separate web AI features remain TrainingDojo-hosted.

ChatGPT App Listing: Pending Directory Approval

There is no ChatGPT setup action to take yet. TrainingDojo will publish a direct listing link here only after OpenAI approves it. Use the current Responses API path below, Claude, or eligible Gemini Spark in the meantime.

Advanced: Connect With the OpenAI Responses API

Install the OpenAI JavaScript SDK, then create a local .env file containing both server-side credentials (do not commit this file):

npm install openai

# .env
OPENAI_API_KEY=<your-openai-key>
TRAININGDOJO_MCP_KEY=<your-trainingdojo-key>

Save this as responses-example.mjs to attach TrainingDojo as a remote MCP tool:

import OpenAI from "openai";

const client = new OpenAI();

const response = await client.responses.create({
  model: "gpt-5.6", // Use a current Responses API model ID.
  input: "Review my connected training platforms and summarize my last six weeks.",
  tools: [{
    type: "mcp",
    server_label: "trainingdojo",
    server_url: "https://trainingdojo.app/api/mcp",
    authorization: process.env.TRAININGDOJO_MCP_KEY,
    allowed_tools: [
      "get_connected_platforms",
      "get_strava_history",
      "get_trainingpeaks_history",
      "get_intervals_history",
    ],
    require_approval: "never",
  }],
});

console.log(response.output_text);

Run it with Node's environment-file support:

node --env-file=.env responses-example.mjs

Keep the TrainingDojo developer key on your server. Never ship it to browser JavaScript or include it in a public repository. This first example allowlists read-only tools. Before adding any write tool, implement the Responses API MCP approval flow and surface the approval request to the user.

A Useful First Coaching Session

Start with a read-only check before asking the AI to write anything:

Use TrainingDojo to list my connected platforms. Then review my last
six weeks of training and my next 14 days. Summarize volume, intensity,
recovery signals, and the three biggest risks in my current schedule.
Do not change or publish anything yet.

Once the summary looks right, ask for a draft plan, review its dates and workload, and only then approve a publish tool. TrainingDojo labels read, write, and destructive tools; the AI client controls whether and when it asks for approval, so review that client's permission settings before enabling writes.

Why Use the OpenAI API Path

• Your application can allow only the TrainingDojo tools relevant to its workflow.
• Read-only tool annotations can support a lower-friction review flow.
• Write and destructive calls can remain behind explicit user approval.
• You can build a focused endurance assistant instead of relying on copy and paste.

Generate the developer key used above from the Advanced section of MCP setup.