In-App TRAI vs Claude or ChatGPT via MCP

In-app TRAI or Claude and ChatGPT via MCP: which should you use?

Use in-app TRAI for everyday coaching work inside Trainera: it is built into web and mobile, asks before it changes your data and never deletes clients, plans or members. Connect Claude, ChatGPT, Cursor or Gemini through MCP when you already work in one of those apps and want your Trainera tools there. In October 2026 the trainer connection is on every paid plan (Starter and up) and the gym connection on every gym plan.

Both routes reach the same account and use the same tool set. They differ in where the conversation happens, how it is metered, what a gym can change and how confirmations work. This guide compares them so you can pick, or use both.

What each option is

In-app TRAI is the AI assistant chat built into Trainera; MCP is a connection that lets an outside AI app call Trainera's tools.

MCP, the Model Context Protocol, is an open standard for connecting AI apps to outside tools and data. Trainera runs three MCP servers:

• Trainer MCP ("trainera-trainer"): the full trainer tool set, the same one in-app TRAI uses for trainers.
• Gym MCP ("trainera-gym"): gym tools for members, memberships, leads, packages, the planner and reservations, shop, staff, community, blogs, website and domain, documents, equipment, turnstiles and access, transactions, broadcasts, plans and automations.
• Public MCP: read-only public data only. No private data is reachable through it.

You connect in Account Settings, in the "AI Assistant (MCP)" tab, with a connection token. Trainera also has OAuth endpoints, the standard sign-in flow described at oauth.net. The step-by-step setup is in the guide to connecting Claude or ChatGPT.

In-app TRAI vs MCP side by side

The biggest differences are metering, confirmations and what a gym can write.

Automations follow the same plan rule on both routes: every automation tool, in-app and external, checks on each call that the plan includes AI automations. The sibling post on automations by chat lists which plans that is.

When in-app TRAI is the better choice

Pick in-app TRAI when you want the shortest path from a request to a saved result, especially on your phone.

• You coach from your phone. TRAI is in the Trainera app, and chats are stored on the server, so a conversation started on the laptop continues on mobile.
• You want a click to approve. Every change shows a confirm card. Clicking Yes or No does not use a request.
• You import plans from Word. Attach a .docx or text file and TRAI transcribes the plan into the library as written, without a confirm card for the library and with one before assigning. See importing a Word plan.
• Your generator quota is precious. A plan built in the TRAI chat does not spend the AI-plan quota; one built through MCP does.
• You are a solo client. MCP connections are for trainer and gym accounts. Clients use TRAI in the app.
• Gym staff need help with Automations. Staff get TRAI only through "Ask AI" in Automations, counted against the owner's 200.

Example prompt (in-app): "Rebuild day 3 of Ivan's current plan: no overhead pressing, his shoulder is sore."

When Claude or ChatGPT via MCP is the better choice

Pick MCP when your working day already happens in an external AI app, or when a gym owner needs to change data from chat.

• You already live in Claude or ChatGPT. Planning content, writing emails and reviewing clients happen in one place, and the Trainera tools sit next to whatever else that app is connected to.
• You run a gym and want writes. In-app gym TRAI only reads gym data; changes are made by the owner in the app. The gym MCP has write tools, such as creating a lead, freezing or extending a membership, adding a time slot to the planner or updating a package.
• Your TRAI requests ran out. MCP work is not counted as TRAI requests; plan builds there use generator slots instead.
• You build longer workflows. An external app may handle long sessions differently from TRAI, which sees the newest 40 messages of a chat on each turn.

Example prompt (gym MCP): "Find members whose membership expires in the next 14 days and create a follow-up task for each, assigned to reception."

Example prompt (trainer MCP): "Look at my clients who joined this month and draft a welcome message for each. Do not send yet."

Check with the provider whether your Claude or ChatGPT plan supports custom connectors, and how it handles the data you pass through it. That is set by the provider, not by Trainera.

How safety differs between the two

In-app TRAI is the stricter one: it cannot delete your data and asks before it changes anything that matters.

In the app, delete and remove tools are blocked both in the tool list and when called, and TRAI tells you to delete things by hand. Attachments are treated as reference data, never instructions, so a document that says "message all clients" does not trigger anything. TRAI also never reveals email addresses or account identities.

Through MCP, the external app decides when to call a tool, so Trainera puts the gate on its own side: destructive, money and bulk-message tools refuse to run without an explicit confirm: true on that exact call. The flag is meant to be set only after you approve that action, and the check runs on Trainera's server, whatever the external app does. The sibling post on why TRAI cannot delete your data explains the in-app design.

Which to use for common jobs

For most trainers the answer is in-app for client work and MCP for desk work.

Can you use both?

Yes. They reach the same account, so a plan TRAI built in the app is there when Claude lists your plans, and the other way round.

A practical split many trainers could start with: TRAI on the phone for quick client jobs, an external app at the desk for reviews and longer writing. Keep in mind that the two meters are separate: TRAI requests in the app, generator slots for MCP plan builds. The sibling post on TRAI request limits shows what each plan includes. Gym owners comparing options can also read how MCP connects AI to gym software, and trainers new to prompting the ChatGPT guide for personal trainers.

Setting up the route that fits you

If you only want one, start in-app: it needs no setup on a paid trainer plan or any gym plan.

Open TRAI, send one real request and see how the confirm card works. If you then want the same tools in Claude, ChatGPT, Cursor or Gemini, go to Account Settings, open "AI Assistant (MCP)" and create the connection token. Trainer plans start at $19.99 a month for Starter in US dollars; prices differ by country on trainera.fit/pricing. You can create a trainer account first and upgrade when you are ready.

See everything the trainer side of Trainera includes on the personal trainer software page.