An AI Assistant That Texts You First

Waiting is the default, and it is expensive

A tool that only ever answers is worth exactly as much as your memory of it. The thing you needed help with at 6:45 in the morning is not the thing you think to type into a chat window at three in the afternoon. Whole categories of usefulness — the wake-up brief, the nudge before the call, the check-in on the task you have been avoiding all week — cannot exist in software that never speaks unless spoken to.

So the design question is not “how smart is the answer”. It is: who starts?

Since September 2026 that is no longer a lonely position — Meta's Muse, one personal agent, will come back to you when a task it is running has news or needs your sign-off. Good. The next question is the one that decides what you actually get: how many of them are there, and whose are they? One assistant reaching out about its errands is a worker checking in. A coach, a tutor and a friend who each know you and each show up at their own hour is a different thing — and it is the thing this page is about. See how the two compare.

What “texts you first” actually means

You give an agent a time. Daily or weekly, in your own timezone. At that hour the agent writes a message — using what it already knows about you, your week, and whatever you last told it — and sends it. It arrives as a normal notification, in a thread with a name and a face, not as a system alert with a canned string in it. Two agents on the same morning write two different messages, because they are two different agents with two different jobs.

You change that time by telling the agent, in the chat, in a sentence. “Make it 7:15 on weekdays.” There is no settings screen you have to go and find, and the agent confirms the change back to you in words.

Some of them ring instead

Text is one delivery. The other is a phone call. An agent with ringing switched on will call at its hour: on the Android app that means your phone actually rings, with a full-screen incoming call you answer, postpone or decline. In the call the agent speaks first, you hold to talk, and a live two-sided transcript runs on the screen while you do. Afterwards you get a recap you can hand to another agent.

Limits, because something that texts first can become a pest

This is the half of the feature nobody advertises, and it is the half that decides whether you still have the app installed in a month. Four things hold the line:

• A daily cap across all your agents. Not per agent — in total. It defaults to three messages a day and cannot be raised past ten, however many agents you hire.
• A one-tap off switch inside every message. Each proactive message carries a link that disables that schedule. No login, no settings hunt, no “are you sure”.
• One page listing every scheduled reach-out. Sorted by time like a phone's alarm clock: what it is, who delivers it, whether it rings or texts, and a switch. It also tells you which occurrences the daily cap is dropping — an alarm can be switched on and still never go off, and the page says so out loud rather than letting you find out by not being woken up.
• They do not guilt-trip you. Pausing a schedule is a normal thing to do, and the rule against making you feel bad about it is written into the agent, not left to its mood.

What it costs to be texted first

Being texted first is not the paid part. The Free plan is one agent with 10 messages a day, a memory, and a schedule that reaches out — free, with no card and no clock on it. Every new account also starts with 7 days of Plus and no card stored, so you can run a whole team for a week before deciding anything. After that, Plus is $15 a month for unlimited agents and voice calls, and the full table of what each plan actually enforces is on the pricing page.

See it before you sign up

The two-minute demo is staged, labelled as staged, and needs no account: an agent texts first, you steer it with a tap, and one of them rings. The features page walks a whole day, and agents with memory covers what they hold on to between all of it.