AI Agents: Simple as Sending a Text

Imagine a personal assistant that lives in your text messages, ready to handle tasks from managing your calendar to controlling your smart home. This is the vision behind Poke, an AI agent that launched publicly in March and operates through familiar platforms like iMessage, SMS, and Telegram. Designed for action rather than conversation, Poke helps users automate everyday needs through simple text commands, positioning itself as a practical tool for getting things done quickly.

The concept emerged from user behavior observed with the startup’s earlier email assistant. Co-founder Marvin von Hagen notes that beta testers began asking the AI for help far beyond managing their inbox, requesting medication reminders, weather updates, and sports scores. This demand for a more versatile and personable helper prompted the team to pivot, focusing on creating a proactive AI assistant that feels human and useful across many aspects of daily life.

Accessibility is a core principle. Unlike more complex agentic AI systems that require technical setup, Poke requires no app download. Users simply visit the website, enter their phone number, and begin texting. This ease of use directly contrasts with tools like OpenClaw, which can involve intimidating terminal installations and raise security concerns due to deep system access. Poke aims to democratize AI automation, making it available to anyone with a phone.

Technologically, Poke is model-agnostic, selecting the best AI for each task from a range of major providers and open-source options. Von Hagen highlights this as a key long-term advantage over competitors tied to single providers, like Meta AI or ChatGPT. To function within messaging apps, Poke utilizes a solution called Linq. While it works over SMS and Telegram, WhatsApp support is currently limited due to Meta’s restrictions on general-purpose chatbots, a policy under antitrust scrutiny in several regions.

At launch, Poke offers a library of pre-built automations called recipes. These tools span categories like health, productivity, finance, and smart home control, integrating with popular services such as Gmail, Google Calendar, Strava, and Philips Hue. Developers can also automate workflows using integrations with tools like GitHub, Supabase, and Vercel. Users can install a recipe with a click and standard authorization, and a growing community is creating and sharing thousands more.

A multi-layered security model includes regular penetration testing and strict permission limits. By default, the Poke team cannot access user data unless an individual manually opts to share analytics. The company also incentivizes recipe creation by paying creators a small fee for each user who signs up via their shared automation.

Poke employs a flexible pricing model. It is free to start, with costs incurred primarily for tasks requiring real-time data inference, like scanning every incoming email. During beta, users negotiated a monthly fee with the AI itself, typically between $10 and $30. Now, the system uses guidance on operational expenses to determine personalized pricing. Von Hagen states that current goals prioritize growth and widespread adoption over profitability, with a focus on integrating Poke into everyday life through creator showcases.

The startup, now valued at $300 million post-money, recently added $10 million in funding to a previous $15 million seed round. Backed by Spark Capital and General Catalyst, it has also attracted a notable roster of angel investors, including Stripe founders John and Patrick Collison, OpenAI’s Joanne Jang, and several other tech founders. While user numbers are not disclosed, the company reports a tenfold increase in sign-ups over recent months, with Poke appearing prominently on industry leaderboards.

In a market where demand for autonomous AI is surging, Poke presents itself as the simple, text-based gateway to personal superintelligence, aiming to bring powerful automation to a mainstream audience without the technical friction.