Preface
For a decade, launching a consumer product meant shipping an app and then fighting for every install. Some teams building consumer AI now skip that step entirely. Their product is a contact you text.
We see this from the infrastructure side. Ditto and Flip, both built on Photon, have no app at all: you text them. Here is what teams like theirs get right, what it takes, and when an app is still the better call.
The install is the step you can skip
Every app asks for the same things before it helps anyone: find it in the store, download it, create an account, allow notifications, learn a new interface. An agent in iMessage asks for one thing: send a text.
Ditto, an AI matchmaker for college students, chose iMessage as its only channel. Onboarding is a conversation, not a form: a student texts a code, Ditto asks what they care about, and every Wednesday at 7 p.m. it sends one match with a planned date. More than 205,000 students have signed up, all over iMessage.
Flip, a finance assistant, works the same way. You get a text and start right away, with no app and no new account.
Sharing is just a text
When the product is a phone number, telling a friend about it is one more message in a thread you already have open. About a quarter of Ditto's growth comes from students bringing in friends.
Group chats go further. Add Flip to the chat for a trip and ask it to split the Airbnb: it reads the thread, does the math and sends each person their share. The agent joins the place where the decision is already being made, instead of asking everyone to move somewhere else.
The conversation is the interface
Messages are where people already coordinate plans, money and work, and they check them all day. Slashy, an email agent, texts you when something needs you: a drafted reply you approve by answering “send”, a meeting that moved, a brief before a call. Flip reports two to three times the retention of classic finance apps.
A message that reads like it came from a person also gets answered. In our own outreach over iMessage, we see more than 60% better response than with traditional SMS (how we research leads before texting them).
What iMessage-first takes
iMessage was not built for agents. The common workaround bridges it from a personal Mac that has to stay awake, sending from someone's own number. That holds up for a side project, not for a product people rely on. An agent that feels native in iMessage needs:
A dedicated line, so it shows up as a real contact with blue bubbles.
Typing indicators, tapbacks, threaded replies and group chats, so it behaves like everyone else in the thread.
RCS and SMS fallback, so Android users and older devices still get the message.
A way to show more than text when text isn't enough. Live mini app cards render an interactive card right in the transcript and fall back to a link where they can't.
That is the layer Photon provides, through one API that also covers WhatsApp, Telegram, Slack and more.
When an app still wins
iMessage-first doesn't have to mean iMessage-only. Some products need long sessions, dense visual browsing or controls that don't fit in a thread, and an app is still the right home for those. A good pattern is to start in messages, learn what people actually ask for, and then decide what deserves a screen of its own. Because the same agent can answer on other channels, adding WhatsApp or Telegram later doesn't mean starting over.
Getting started
If you are building a consumer agent, How to build an iMessage agent in 2026 compares every approach, and How to choose an iMessage API in 2026 covers what to look for in a provider. To try Photon, start with the docs.