Is connecting ChatGPT to ReminderIt safe?

Giving an AI assistant the ability to place real phone calls is a fair thing to be cautious about. The good news is that the ReminderIt connection is built around consent you grant explicitly and can withdraw at any time, and it is scoped to your account alone.

Here is exactly how the connection works, what ChatGPT can see, and how you stay in control.

You approve access through OAuth

Connecting ReminderIt to ChatGPT uses OAuth. You add the hosted connector at https://mcp.reminderit.com and are taken to a secure ReminderIt sign-in screen where you approve access. Nothing connects until you complete that step yourself.

There is no API key to copy and paste when you use the hosted connector, which means no secret sitting in your clipboard or a config file for you to look after.

Per-client, revocable keys

Behind the scenes, approving access creates a key specific to that client in your account. Because it is per-client, connecting ChatGPT and connecting another assistant produce separate keys, and revoking one does not disturb the other.

You can see and revoke these at reminderit.com/api-keys. Revoking is the clean off-switch: the moment you revoke, that client can no longer act in your account.

Scoped to your account only

The connection acts strictly within your own ReminderIt account. ChatGPT can see and change your reminders, your voice settings and your usage; it cannot reach anyone else's account, and no one else's assistant can reach yours.

Adding another person as a care-recipient is deliberately gated: add_recipient triggers a consent call to that person before any reminder can go to them. You cannot quietly schedule calls to someone who has not agreed.

What data is actually visible

Through the tools, ChatGPT can read what it needs to help: your list of reminders and their details, your upcoming calls, your recent call history via get_recent_calls, your plan and time zone via whoami, and your quota via get_usage. It works with these to schedule and adjust calls.

That visibility is the same information you see in your own dashboard. The assistant is not given a wider view than you have.

Staying in control

A sensible routine: connect when you want the convenience, use test_call to confirm things behave, and revoke at reminderit.com/api-keys if you stop using a given assistant. Because keys are per-client, you can keep the ones you use and cut the ones you do not.

In short, the connection is explicit to grant, narrow in scope and instant to revoke. That combination is what makes handing an assistant the ability to call your phone a reasonable thing to do.