AI Agents Now Live in Your Texts. Should Your Product?
TechCrunch just listed more than 20 AI agents you can text like a friend, from Poke to Instinct, now valued at $10 billion. Texting removes the app download, but it brings platform rules, message fees and new costs. Here is how to decide if your product belongs in the chat.

On November 21, 2026, TechCrunch published a list of AI agents that "live in your text messages." It runs to more than 20 products. You text them like a person, and they remember context, connect to your apps and get things done.
Some are tiny betas. Some are not. Instinct, which gives assistants their own email addresses and can make phone calls, raised $1 billion at a $10 billion valuation in September, according to the list. Tab came out of stealth in October at a $300 million valuation. Poke, the first AI agent Apple approved for its Messages for Business platform, was bought by Cognition in July.
If you are building any product that answers questions, books things or nudges people to act, you have probably wondered: should this just be a phone number people text? Here is how to think it through.
Why Texting Is Suddenly a Product Channel
The pitch is simple. People already spend their day in messages. An agent that lives there needs no app download, no new login and no habit change. You save a contact and start typing.
The TechCrunch list shows how wide this has spread:
- Personal assistants like Poke, Tomo and Maverick work in iMessage. Tomo can join group chats. Its basic plan starts at $19.99 a month.
- Family organizers like Ohai and Orbits handle household schedules. Ohai has a free tier and paid plans from $9.99 a month.
- Work tools like Ask Proxy run over SMS, with business pricing of $15 to $23 per employee per month.
- Niche agents target travel (Miso, which pairs AI with a human travel team), creators (Stanley) and small businesses (Szn, which claims more than 7,500 businesses served in 16 countries).
Notice the pattern. Most of these are not new ideas. Calendar help, reminders and travel booking all exist as apps. The bet is that the channel itself is the product advantage.
What Poke Teaches About the Upside
Poke is the clearest case study. It launched publicly in March 2026. In June, AppleInsider reported that Apple approved it as the first third-party AI agent on Messages for Business, the system companies use to chat with customers inside iMessage.
By July, TechCrunch reported Poke had "hundreds of thousands" of users and more than 100 million messages exchanged over the prior three months. The most common use was plain productivity: email, reminders and to-dos. Cognition, the maker of the Devin coding agent, bought Poke's parent company in a deal valued in the low nine figures. The stated reason was not the channel. It was Poke's personality and interaction style, which Cognition wanted for its own product.
That is a useful lesson. In a chat window, your product has no buttons, menus or design system to lean on. Its tone, memory and judgment are the interface. If people love talking to it, that is a real asset.
What Poke Teaches About the Downside
The same TechCrunch report said Poke had been expensive to run and hard to make profitable. That is the part founders tend to skip.
Every message costs you something. You pay for the AI model to read and reply. You may pay the platform to deliver the message. And because users text casually, they send a lot of messages, many of them low value. "Thanks!" still triggers a model call unless you handle it.
AppleInsider also noted that Apple charges Poke's maker per user on its platform, and that Poke's own pricing was set "through negotiation" with users, with light tasks free and heavy ones paid. That is a sign of a company still figuring out how to charge for something people expect to be as cheap as texting a friend.
Your AI co-founder is ready when you are.
Foundra turns everything in this article into an actual plan. Validation, customers, pricing, launch. In one place, in your voice, in an afternoon.
Get started→$39/month. Cancel anytime.
The Platform Rules Are Real
When your product lives inside someone else's app, that company sets the rules. Meta's WhatsApp is the clearest example.
According to Meta's developer documentation, WhatsApp moved to per-message billing on July 1, 2025. Businesses pay when a message is delivered. Free-form replies, called service messages, can only be sent inside a 24-hour window after the user writes in. Since October 1, 2026, those service messages cost money after the first 1,000 per phone number each month.
The rules for AI companies are stricter. Under terms updated January 15, 2026, Meta allows general-purpose AI assistants on the WhatsApp Business Platform only where it is legally required to. In those markets it has charged AI providers per service message, with dates that differ by country. A specialized business bot is treated differently from a general chatbot.
The takeaway: a channel you do not own can change price or policy with one update. Build so you could survive if it did.
Should Your Product Live in the Chat?
Texting is a strong fit when most of these are true:
- The job is short and frequent. Reminders, quick answers, approvals and status checks work well. Long tasks with lots of options do not.
- The user would otherwise forget you. A text can reach someone who would never open your app. Proactive follow-ups are the real superpower.
- The answer is text, not a screen. If users need to compare ten options, see a chart or edit a document, a chat window will frustrate them.
- You can afford the per-message cost. Estimate messages per user per month, multiply by model and delivery costs, and compare it to what each user pays.
- You can handle trust. People text sensitive things. You need clear consent, data handling and a way to stop messages instantly.
If only one or two are true, consider texting as a feature, not the product. Send reminders and confirmations by text, and keep the core experience in a web app you control. This is also a good moment to sketch your go-to-market plan on Foundra or on paper, so you can compare the chat channel against the channels you already own.
How to Test It Cheaply
You do not need Apple approval or a big build to test demand.
- Start with a "concierge" number. Give 20 target users a phone number to text. Answer by hand, or with a simple AI draft you review, for two weeks. Track how often they come back.
- Measure messages per task. If it takes 15 back-and-forth texts to book one thing, your costs will balloon and users will get tired.
- Price early. Ask testers to pay a small monthly fee after the trial. Free usage tells you little about a product that costs money per message.
- Pick one channel first. SMS reaches almost everyone. iMessage is great on iPhones. WhatsApp dominates in many countries. Each has different rules and costs, so start where your users already are.
- Plan for a web fallback. Keep a simple web page where users can see history, settings and billing.
The Bigger Picture
The rush into text-based agents is part of a wider shift. TechCrunch also noted privacy and security worries about agents that act on their own, especially ones that can spend money or make calls. Expect platforms and regulators to tighten rules, not loosen them.
For a first-time founder, the opportunity is real but narrow. The winners will not be the ones who simply moved an app into iMessage. They will be the ones whose agent is truly pleasant to talk to, does a small job reliably and costs less to run than users pay. Get those three right, and the channel works for you.
FAQ
What are AI agents in text messages? They are AI assistants you reach by texting a phone number or a contact in iMessage, SMS, WhatsApp or other chat apps. They keep context, connect to other apps and can complete tasks.
Which companies are building them? TechCrunch's November 21, 2026 list includes Poke, Tab, Instinct, Tomo, Maverick, Ohai, Orbits, Ask Proxy, Martin, Town and others.
Is it expensive to run a texting agent? It can be. You pay for AI model calls on every reply and often for message delivery. TechCrunch reported Poke was expensive to run and hard to make profitable.
Can I build an AI chatbot on WhatsApp? Business bots are allowed under Meta's rules, but general-purpose AI assistants are only permitted where Meta is legally required to allow them, under terms updated January 15, 2026.
How much does WhatsApp charge businesses? Meta charges per delivered message by category and country. Since October 1, 2026, service messages are charged after 1,000 free per phone number each month.
Should my startup be text-first? Only if the job is short, frequent and text-friendly, and your price covers per-message costs. Otherwise use texting as a feature.
Sources
- Here are the top AI agents that can live in your text messages, TechCrunch, November 21, 2026
- First AI agent for Messages business chat approved by Apple, AppleInsider, June 4, 2026
- Why Cognition bought Poke: AI personality is becoming a competitive advantage, TechCrunch, July 24, 2026
- Pricing on the WhatsApp Business Platform, Meta for Developers, effective October 1, 2026
- New pricing policy for AI Providers leveraging the WhatsApp Business Platform, Meta for Developers, updated October 15, 2026
You just read the theory. Ready to build the thing?
Foundra is your AI co-founder. It turns an idea into a validated business plan, a go-to-market, and your first 10 customers. In an afternoon, not a semester.
$39/month. Cancel anytime. Works in 20 languages.