AI Agents for Text Messages: 24 Tools to Know
Explore 24 AI agents that work through text, compare family, work and travel tools, and use a practical checklist to choose one without ignoring privacy.
The AI Agents Moving Into Your Text Messages
The next AI assistant you try might not ask you to download an app. It could show up as a text conversation - asking for a reminder, pulling up a restaurant, or helping you wrangle a busy week. The big shift is that using AI isn't only about getting an answer. It's about moving the work from your inbox and chats to wherever the task actually gets done.
That's what a text-based agent is for. Some can remember preferences, connect to other apps, and take actions like scheduling or sending a message. Others stay focused on one narrower job - family logistics, travel planning, or helping a creator prep posts. This category is still young, and a familiar messaging interface doesn't automatically mean the agent is safe or dependable.
This guide compares 24 notable options, explains what separates a chatbot from a task-taking agent, and walks you through how to choose without handing over too much. Availability and pricing can change; the details here reflect what was available in the source material as of October 2026.
What makes a text-based AI agent different
A conventional chatbot responds to prompts. An agent is meant to carry work forward: it may connect to a calendar, email account, browser, or messaging service, then carry out steps and report back. In practice, the difference is less about the label and more about access - what the product can reach, what actions it can take, and whether it can confirm the task is actually finished.
Text works as an interface because it's already part of daily routines. You can forward an email, dictate a request, or ask a question without switching apps. For example, someone might send a school notice and ask an assistant to add the event to a shared calendar. At work, an agent might summarize a call and surface follow-ups later.
Of course, convenience comes with trade-offs. The agent needs context, integrations, and permission to act. If any of that is shaky, the whole experience can wobble - misreading a request, failing to reach the right service, or making a change you didn't mean to authorize. When you evaluate one, don't just judge how smooth the conversation feels.
From answering to completing
A useful way to assess an agent is to split its capabilities into three levels:
- Information: It can answer questions or summarize what you provide.
- Coordination: It can connect tools, create reminders, or lay out next steps.
- Action: It can make bookings, send messages, change appointments, or otherwise do things on your behalf.
That third level is where the value gets real - and where mistakes can get expensive. Drafting a message isn't the same as sending it. Proposing a calendar change isn't the same as actually changing your schedule.
Messaging does not mean no setup
These tools may work over SMS, iMessage, RCS, WhatsApp, Telegram, Signal, or another channel. Some still require an app, account connections, or a subscription. Before you pick, check which messaging options work on your devices, what integrations are available where you live, and whether the product is in beta or private testing.
And not everything in this space lives strictly inside a chat. Skye, for instance, is described as an iPhone homescreen assistant that shows cards and briefings. Underdog's first release runs locally on computers rather than operating like a messaging agent. They're relevant to where personal AI is heading, but they aren't direct substitutes for a text-first assistant.
24 AI agents, grouped by use case
The easiest way to compare this market is by the job you're trying to get done - not by the word "agent." A personal assistant, a household coordinator, and a travel tool solve different problems, even if their feature lists overlap.
General-purpose personal assistants
- Ask Proxy works over SMS for everyday decisions, gift ideas, event planning, reminders, and recommendations, and it also has a work-focused version. The company describes it as proactive, including reaching out with reminders or suggestions. The personal plan is listed at $1 per day; business pricing varies by team size.
- Caddy turns details from messages and email into actions, including calendar events, reminders, and follow-ups. It works through iMessage on iPhone and RCS on Android and was in public beta during the source period.
- Comma handles work and personal tasks across a user's browser, computer, and files, with access through services including Signal, Telegram, and WeChat. It emphasizes carrying tasks through to completion, keeping context across devices, and flagging work that needs review. It's free to start and open source.
- Folk connects via iMessage, WhatsApp, and Telegram. It can handle reminders, email, flight tracking, research, and bookings. Its private cloud computer is meant for multi-step tasks. A free option is available, with a Pro plan listed at $8.33 per month for unlimited background tasks.
- Instinct is a private-beta assistant you can contact by text or voice. It connects to services like email, calendars, and Google Workspace. The company says users have delegated tasks such as trip planning, grocery shopping, and subscription cancellations. It has also introduced dedicated email addresses for assistants and has begun rolling out calling support. Because of that, permission and account security matter a lot.
- Iris uses the Hermes Agent system and works through iMessage. It's designed to keep relevant context, handle tasks in connected apps, and follow up proactively.
- Martin is reachable through SMS, phone, WhatsApp, email, Slack, and an iOS app. It can help with calendars, email, reminders, and communications, including sending messages or making calls. Subscriptions start at $21 per month.
- Maverick works through iMessage and can help with tasks like fixing typos, scheduling meetings from voice notes, and coordinating dates in group chats. It says it asks permission before taking action. The beta is free; a Pro plan is listed at €27 per month.
- Pally connects information across services like WhatsApp, Gmail, calendars, and Google Drive, with access through iMessage, RCS, and Telegram. It has a free plan.
- Poke is a general assistant for calendars, planning, health and fitness, smart-home controls, and photos. It launched in 2026 and became available through Apple Messages for Business.
- Rene turns natural-language conversations into tasks, reminders, and summaries. It's also described as having a browser plus the ability to write code, shop, and create slides and images. It's available on iMessage, Telegram, and WhatsApp.
- Tab lets users delegate everyday requests through iMessage or WhatsApp, including grocery orders and gift searches. It came out of stealth in October 2026; the source describes a $300 million valuation, not a consumer subscription price.
- Tomo is designed for both group conversations and one-on-one chats. It supports organization, scheduling, reminders, web searches, and work with calendars and email. Its basic plan starts at $19.99 per month.
- Wajo's Fo can call businesses, send emails, join group chats, and make bookings or purchases. Fo has its own email address, phone number, and payment card. The company also says a human assistant can step in when the agent can't complete a task alone. A free tier is available, with a paid Pro offer for heavier use and custom agents.
Family and household coordination
Family assistants solve a specific operational problem: useful details are scattered across school emails, calendars, activity schedules, and group chats, but someone still has to turn it into an actual plan.
- Fambot combines school communications, activities, meals, calendars, and household tasks. It works through an app, web, and SMS, and it sends a daily summary of the following day. It connects to Gmail, Google Calendar, Outlook, and WhatsApp, with Apple Calendar support planned. The beta was free, with a future price expected to be around the cost of a Netflix subscription.
- Ohai turns messages, forwarded emails, and voice requests into schedules, reminders, meal plans, and household coordination. It has a free basic option, with paid plans starting at $9.99 per month.
- Ollie brings together family calendars, emails, tasks, and reminders, then sends relevant details by text. It offers a free tier and paid plans starting at $25 per month. Its reported SOC 2 compliance may matter if you're comparing how providers handle security, though compliance alone doesn't tell you whether it's a good fit for your family.
- Orbits combines an app and text messaging to coordinate calendars, lists, chores, and service-provider communication. It can also search the web, make purchases, book restaurants, change appointments, and request household-service quotes.
For a household, the real test is whether multiple people can use it without adding extra coordination work. Check who can see a reminder, who can approve a purchase, and how the assistant handles conflicting calendars.
Specialist and adjacent agents
- Miso focuses on travel through iMessage, combining AI trip planning with support from a dedicated travel team. It takes preferences and loyalty points into account and can help arrange flights, provide updates, and make recommendations.
- Szn coordinates tasks and follow-ups through a conversation thread. Users can send text, voice notes, photos, or forwarded emails; the service can make calls, send emails, and manage follow-ups. The company reports serving businesses across 16 countries, which makes it more business-oriented than a personal assistant.
- Stanley is aimed at creators. It can develop content ideas, turn voice notes into scripts and captions, connect to Instagram for context, and support a publishing calendar and post scheduling.
- Town focuses on professional work, with a Townie assistant designed to learn recurring workflows across email, calendars, documents, and Slack.
- Skye is an adjacent iPhone homescreen assistant rather than a text-thread agent. It aims to surface contextual cards, meeting prep, reminders, and audio briefings from connected services. It was in private testing and beta, with no public release date announced in the source period.
- Underdog is a privacy-focused adjacent product. Its invite-only beta runs the model on the user's device, initially on Mac and Windows, with other platforms planned. That local-first approach is different from the cloud and messaging models used by many agents on this list.
How to choose an agent without handing over too much
Don't pick the assistant with the biggest feature list. Pick the one that can handle a recurring task you already deal with, using access and permissions you're comfortable granting.
Use this five-part evaluation
- Start with a repeated task. Choose something specific, like pulling appointment details from messages or assembling a daily family schedule. A tool that solves one recurring problem is easier to judge than one that promises to handle everything.
- Check the full workflow. Ask what happens from request to completion. Does it only suggest a calendar event, or can it actually create one? Does it show you the proposed action? Can you correct or cancel it?
- Map the integrations. Confirm the services you rely on are supported. If the agent can't access your real calendar or messaging tools, it may just create another inbox to manage.
- Review the permission boundary. Find out what data it connects to, whether it can send messages or make purchases, and how you revoke access. Dedicated email addresses, payment methods, and calling features should be treated as meaningful access - not minor conveniences.
- Calculate the real cost. Compare subscription price with the time saved and the work you'll still need to double-check. Look at message limits, task limits, background execution, and whether it's only free during beta.
Keep human approval where mistakes are expensive
For low-risk tasks - like drafting a packing list - automation can be fine. But for payments, bookings, account creation, or messages sent in your name, approval checkpoints are usually worth the extra step. A fast workflow isn't "efficient" if you later have to undo an incorrect purchase or explain an unintended message.
Privacy and security deserve the same scrutiny as capability. Ask how long conversation data is retained, whether connected content is used to improve models, and what happens when you delete your account. If the provider doesn't explain the basics clearly, don't connect your most sensitive accounts just to try a demo.
Here's the trade-off in plain terms: more access can mean more useful actions, but it also means a bigger blast radius if something goes wrong. Start with limited permissions, verify what the agent does, and expand access only after it earns your trust.
Conclusion
Text-based AI agents are making a straightforward bet: people are more willing to delegate routine work when the assistant is available inside the conversations they already use. That's a smart interface decision, but it doesn't magically solve the hard parts - context, permissions, reliability, and accountability.
Choose based on the job you need done, not on the breadth of promises. Start with one repeatable workflow, keep approval for high-impact actions, and track whether it actually reduces coordination work. If it does, expand carefully. If it mostly becomes another system you have to watch, then the text interface hasn't fixed the real problem.