The AI Agents You Can Just Text: A Field Guide to the New Assistant Stack

There's a quiet architectural shift happening in consumer AI, and it's easy to miss because it doesn't look like a product launch. It looks like a text thread.

Instead of building another app with another login and another notification badge, a growing crop of startups are betting that the interface you already use fifty times a day, your messages app, is the only interface that matters. You text the agent what you need. It remembers context across conversations, connects to the calendars and inboxes you already run your life through, and actually executes: booking a reservation, rescheduling a meeting, researching a flight, nudging you about something three days from now. No onboarding flow required.

The buzziest name right now is Instinct, which went from a $350 million raise at a $2.5 billion valuation in August to a $1 billion Series C at a $10 billion valuation just a month later in September. That kind of valuation jump for a product still in private beta tells you how much capital is chasing this category. But Instinct is far from alone, and the approaches diverge in interesting ways once you look past the headline numbers.

The Field: General Assistants, Family Tools, and Niche Players

Most of these products cluster into a few recognizable buckets, and the differences come down to what they're optimized to do with the context they collect.

General-purpose assistants are the most crowded lane:

• Caddy lives natively in iMessage (iPhone) and RCS (Android), parsing emails and conversations for things that need action, like adding a calendar event buried in a message thread. It's been in public beta since April 2026.
• Comma is free and open source, works across browser, computer, and files, and stays reachable via Signal, Telegram, and WeChat. It checks its own work before surfacing completed tasks for approval, which is a meaningfully different trust model than agents that just report back and move on.
• Folk runs on its own private cloud computer rather than a shared backend, and can execute code and multi-step tasks, not just respond conversationally. It's reachable over iMessage, WhatsApp, and Telegram, free to start, with a $8.33/month Pro tier for unlimited background tasks.
• Iris, powered by Hermes Agent, works specifically through iMessage and is built to follow up proactively rather than wait for the next prompt.
• Martin is the most channel-flexible of the bunch, reachable via SMS, phone, WhatsApp, email, Slack, and its own iOS app. It's also the priciest, starting at $21/month.
• Pally leans into a conversational, friend-like tone and bundles in actual phone call minutes, 15 minutes free, 30 minutes for $25/month, 60 for $100/month.
• Poke launched in March 2026 and became the first AI agent approved on Apple's Messages for Business platform in June. A month later its parent company, The Interaction Company of California, was acquired by Cognition.

Then there's a cluster built specifically for household logistics, which makes sense given how much of family coordination already happens by text anyway:

• Fambot sends a nightly SMS summary of the next day's events, to-dos, and logistics (school uniform requirements, packing lists), and parents can just reply to adjust things. It connects to Gmail, Google Calendar, Outlook, and WhatsApp, with Apple Calendar support planned. It raised $3.5 million in pre-seed funding and launched in beta in early September 2026.
• Ohai turns forwarded emails, texts, and voice requests into shared schedules and chore/meal coordination, with a free tier and paid plans from $9.99/month based on family size.
• Ollie is notable less for features and more for compliance: it's one of the first mainstream family-focused AI assistants to achieve SOC 2 certification, a security standard that matters a lot once you're handing an agent access to your kids' school emails and your shared calendar. Plans start at $25/month for 150 messages, scaling to $100/month for 1,000.
• Orbits extends past the immediate household, letting the agent coordinate with external service providers and request quotes, not just manage internal family logistics. It's backed by a16z's Speedrun fund, N49P, and Garage Capital.

And then there's Miso, which narrows the scope entirely to travel, pairing iMessage-based trip planning with an actual human travel team backing up the AI, taking into account loyalty points and preferences when booking flights.

Key Takeaways

The common thread across all of these is that messaging apps are becoming a legitimate agent runtime, not just a notification channel. That's a real infrastructure bet: it means these companies are building integration layers into iMessage, RCS, SMS, WhatsApp, Telegram, and Signal rather than investing in native app UX, which changes what "distribution" even means for an AI product.

It also raises the stakes on trust and permissions. Instinct's move to give users dedicated email addresses for their assistants, so the agent can sign up for services and manage accounts without cluttering a personal inbox, is a clever bit of account isolation. But it's also exactly the kind of autonomy that's already drawing privacy and security scrutiny, since an agent with its own email and the ability to place phone calls is functionally an unsupervised employee with access to your accounts.

Pricing models are still all over the place, from free-and-open-source (Comma) to $100/month tiers gated by message or call minutes (Ollie, Pally). Nobody's settled on whether this should be priced like a subscription, a utility, or a labor substitute. Given the valuations flowing into this space, that pricing question is probably the next thing worth watching closely.