TechCrunch rounds up AI agents that live in text messages, from personal assistants to family and work helpers

TechCrunch has compiled a list of the most notable AI agents that work directly inside text messages and split it into four categories: general-purpose assistants, agents for families, for travel and for work. The announcement does not reveal the full list, but the feature itself shows a shift in interface. An agent no longer has to be a separate app or a browser tab. A phone number is enough.

Why agents are moving into messaging apps

Chat has three advantages that are hard to argue with. First, there is no barrier to entry. You don't need to download a client, sign up or get used to a new screen. You just add a contact and write. Second, messages are a natural fit for asynchronous work. An agent can take a task, go off to complete it, and then message you on its own when it's done. Third, chat is already built into our social fabric. You can bring an agent into a group conversation with family or colleagues, and it will see the context, not just a single request.

This path did not begin yesterday. As early as 2024–2025, phone and WhatsApp versions of popular chatbots appeared, and startups began building personal agents that talk to the user in an ordinary messenger. TechCrunch's roundup shows that the market has grown out of experiments into a niche of its own, with its own subtypes.

Four categories and what people look for in them

The way the list is divided shows which tasks people hand to agents first:

• General-purpose assistants: answering questions, reminders, message drafts, information search. The main metrics here are response speed and model quality.
• Family agents: a shared calendar, school schedules, shopping lists, arrangements made in a group chat. The value lies in coordinating several people.
• Travel agents: finding tickets and hotels, tracking flight changes, booking confirmations. Integrations with providers and the ability to pay are critical here.
• Work agents: handling email, meetings, tasks and documents. The key requirement in this group is access to corporate data and control over that access.

What's the catch?

Convenience has a price. Ordinary SMS is not protected by end-to-end encryption, which means everything you write to an agent passes through the carrier. Messengers such as WhatsApp encrypt conversations between people, but when the other party is a bot, the data is processed on the developer's side. Then comes the question of trust: an agent that you have given access to your email, calendar and payment details gets broad authority, and incoming messages can become a channel for prompt injection, where outside text tries to make the model carry out someone else's command.

There is also dependence on platforms. The rules for bots in messengers change, and Meta has previously tightened its terms for general-purpose chatbots in the WhatsApp Business API. If an agent lives on someone else's platform, a single policy update can restrict it. Finally, there is the economics: every message costs money, and long sessions with a model cost even more, so many services introduce subscriptions or limits.

How to choose a messaging agent

Before you give an agent access to your life, check a few things:

• Which channel it uses (SMS, RCS, iMessage or WhatsApp) and how the conversation is protected.
• Which permissions it requests: reading email, calendar, contacts, payments.
• Whether you can confirm risky actions with a separate message, for example before buying a ticket.
• What happens to your conversation history: whether it is stored, whether it is used for training, whether you can delete it.
• How much it costs to use and whether there are message limits.

It makes more sense to start with a narrow task such as reminders or a shopping list. If the agent handles it and doesn't ask for unnecessary permissions, you can trust it with more. If its access requirements grow faster than its usefulness, that is a reason to stop.