Skip to main content

MCP | How to Connect Multiple Strety Accounts to One AI Tool

Written by Dan Lapuz

The Strety MCP connector can connect a Strety account to an AI tool account, such as Claude or ChatGPT. This article covers the current process for connecting multiple different Strety client accounts to one AI tool account.

The Strety development team is actively investigating a more official way for Coaches to set up an MCP connection from the Coach Portal in Strety. Because the Coach Portal already has access to a Coach’s client accounts, that future setup would avoid the need to manually connect each individual client account.

In the meantime, the existing MCP setup does allow individual Strety client accounts to be connected manually. Manual setup can be useful, but it comes with added risk. Connecting multiple Strety accounts to one AI tool account increases the chance of pulling data from one client account into another, so each client should stay in its own chat or agent whenever possible.

Use this guide to understand the current manual setup options, choose the safest available connection method, and reduce the risk of mixing data across client accounts.

Warning: By setting up and using manual MCP connections for multiple Strety client accounts, you accept the high risk of cross-account data exposure or data being pulled from one client account into another. Use one client per chat or agent whenever possible.

Below is a written overview of the video above, but we HIGHLY recommend watching the full video above.

Before you start

You may need:

  • Access to each Strety account you want to connect

  • Admin permissions for the Strety account if you are creating a dedicated OAuth app

  • Permission to create custom MCP connectors in Claude or ChatGPT

  • Developer Mode or plugins enabled in ChatGPT if you are setting up a custom MCP there

Connecting the Strety MCP to Multiple Instances — Current Process

Notes pulled from the walkthrough video above. This covers connecting more than one Strety account (instance) to Claude and ChatGPT at the same time. Everything here is still being figured out and is likely to change.

The core problem

The single connector from the directory only lets you point at one account at a time. If you belong to multiple Strety accounts, the authorization step makes you choose which org to connect — and it defaults to your last-visited org (wherever X.strety.com takes you). Nothing warns you if you pick the wrong one, so it's on you to choose the right account.

To actually use two or more instances at once, you have two options below.

Option A — Quick / "hacky" way (community connector + naming trick)

The fast path most people use, but the least safe.

  1. Go to Custom Connectors → Add.

  2. Give the connector a name (e.g. "Demo").

  3. Modify the URL so it's distinct from the established community connector: add a ? followed by a name to the end of the URL (e.g. https://.../mcp?demo). This is what lets the custom connector coexist with the community one instead of colliding with it.

  4. Authorize, and during authorization manually pick the account you intend to connect.

  5. To add another instance, create another custom connector, giving it its own name and its own ?name suffix on the URL. It will then let you create a second instance and connect it to a different account.

Caveats:

  • It uses a generic OAuth app, so it's on you to pick the correct org every time.

  • Nothing stops you from picking the wrong account.

  • No setup work required — click, name, choose account, done.

Option B — Best / "power user" way (dedicated OAuth app per instance)

More setup, but locks each connector to one specific organization so it can never touch another.

  1. In the Strety instance, go to Client Admin → API / Apps (admin permission required).

  2. Create an OAuth app for that org and grab its Client ID and Client Secret to paste into the ChatGPT/Claude side.

  3. When adding the custom app in Strety, choose MCP (not the default) and paste in the Client ID and Secret.

  4. Create and authorize.

Result: the connector is locked to that one organization. No account picker appears — it can only ever connect to that org. Repeat per instance.

Claude vs. ChatGPT

The concept works the same in both, but the mechanics differ.

  • Claude: Any admin who can already create custom MCPs can set these up.

  • ChatGPT: Requires Developer Mode / plugins enabled, plus a publishing step to the org. Permissions are separate from Claude (having access in Claude does not mean you have it in ChatGPT).

  • ChatGPT recommended approach: Create a custom Agent per client and attach the single locked-down MCP to it. You can configure a custom MCP directly inside the agent (Create Agent → Start blank → Browse apps → Custom MCP, which needs Developer Mode). This keeps each client clean and contained. You can add the Client ID/Secret under the agent's Advanced OAuth Settings, same as Option B.

Best practices & warnings for running multiple instances

  • Keep each client contained. The biggest risk is the agent pulling data from one client and writing it into another — there's no way to prevent that, and it's a bad look if a client sees it.

  • One client per chat. Best practice is one instance per conversation; start a fresh chat to switch clients.

  • Read-only is safer than writes. Gathering info carries far less risk than letting it create/update records across accounts.

  • Turn off memory to reduce the chance of cross-account contamination.

  • Use containers: Claude Projects or ChatGPT custom Agents keep context tied to one client. (Note: you currently can't attach a custom app/MCP to a Claude Project directly — agents are the cleaner fit.)

Did this answer your question?