The AI agent race has a new contender, and it doesn't require you to download anything. Poke, a Palo Alto-based startup backed by Spark Capital and General Catalyst, has built a personal AI assistant that lives entirely inside your text messages — accessible through iMessage, SMS, Telegram, and in some markets, WhatsApp.
The concept is deceptively simple. Instead of opening yet another app or navigating a chatbot interface, you just text Poke like you would a friend. Ask it to manage your calendar, track your fitness goals, alert you about specific emails, remind you to take medication, or check the morning weather. It handles the task and texts you back.
The Anti-OpenClaw
Poke arrives at a moment when agentic AI is the hottest category in tech. OpenAI acquired OpenClaw's creator earlier this year, and Nvidia's Jensen Huang declared that every company needs an agentic AI strategy. But for the average person, tools like OpenClaw remain intimidating — they require terminal access, dependency management, and a comfort level with code that most consumers simply don't have.
That's the gap Poke is targeting. Co-founder Marvin von Hagen says the idea crystallized when the team noticed how beta testers were using an earlier email-focused product. People kept asking the AI for things far beyond email — sports scores, medication reminders, weather alerts, daily planning. They liked the conversational feel and wanted it to do everything.
So the team pivoted toward a general-purpose assistant delivered through the most universal interface on the planet: text messaging.
How It Works
Getting started takes seconds. Visit Poke.com, enter your phone number, and start texting. No app install, no account creation flow, no onboarding screens.
Under the hood, Poke routes each request to whichever AI model fits the task best — whether that's a frontier model from a major lab or an open source alternative. Von Hagen sees this model-agnostic approach as a long-term advantage over competitors like Meta AI or ChatGPT, which are locked into their parent companies' models.
Poke also offers "recipes" — pre-built automations that connect to services users already rely on, including Gmail, Google Calendar, Notion, Strava, Oura, Fitbit, Philips Hue, and Sonos, among others. Installing a recipe is a single click plus standard authorization. Developer-focused recipes integrate with tools like GitHub, Vercel, Sentry, and PostHog.
Users can also write their own automations in plain text and share them with others. Thousands of community-created recipes have already been built in just the past few weeks.
The Business Model — Sort Of
Poke's pricing is unusual. During beta, users actually negotiated their monthly price with the AI agent itself, landing somewhere between $10 and $30. Now, pricing is based on usage. Simple requests that don't require real-time data can be free. Resource-intensive tasks — like monitoring every incoming email or running live flight check-ins — cost more, with the AI setting personalized prices based on guidelines from the team.
But profitability isn't the priority. Von Hagen is blunt about that. The goal is growth and daily adoption by as many people as possible. Monetization comes later.
Big Names, Big Ambitions
The startup has raised $25 million total, including a recent $10 million round that values the company at $300 million post-money. Its angel investor list reads like a tech industry yearbook: Stripe founders John and Patrick Collison, Jake and Logan Paul, Dropbox co-founder Arash Ferdowsi, Hugging Face co-founder Thomas Wolf, and executives from OpenAI, DeepMind, and Cognition, among others.
The company isn't sharing user numbers but says signups have grown tenfold in recent months.
The Bigger Picture
Poke's bet is that the future of AI agents isn't about building the most powerful system — it's about meeting people where they already are. Most consumers won't install a command-line tool or learn prompt engineering. But everyone knows how to send a text.
If Poke can make that experience reliable and useful enough to become a daily habit, it won't just be competing with other AI agents. It'll be competing for a slot in the place where people spend more time than almost anywhere else on their phones: the messaging inbox.
That's a crowded space. But it's also the one with the fewest barriers to entry.
