OpenClaw, Devin, Claude Code… AI agents have exploded in 2026, but they're all developer tools. You need a terminal, dependencies, and patience for debugging. Regular people? They're completely left out of this wave. On April 8th, TechCrunch covered Poke — a startup aiming squarely at that gap. It's an app that lets you run AI agents through iMessage, SMS, or Telegram, just like texting a friend. No app install — just go to Poke.com, enter your phone number, and you're done. Spark Capital and General Catalyst are backing it: $25M raised, $300M valuation. CEO Marvin von Hagen put it well — "We don't want to make money. We want to build a product that a billion people use."
What's Actually New Here?
Compare it to other AI tools and the difference is clear. ChatGPT and Claude are "answer my question" tools. Poke is a "gets into my daily life" tool. That sounds like a small distinction, but the use case is completely different. Von Hagen told TechCrunch: "ChatGPT is where you go to ask questions and do research. Poke is where you go when you want to get something done fast, or automate to save time."
"OpenClaw for non-developers" — autonomous agents with no terminal, no dependencies, no error troubleshooting required.
The founding story is telling. The Interaction Company started as a 10-person Palo Alto startup building an email assistant. But beta users started asking Poke to handle everything else — "remind me to take my meds," "text me every morning if I need a jacket today," "send me last night's game score." So they partially pivoted and expanded into a general-purpose assistant. A classic case of user behavior pulling the product forward.
- $300M Valuation
$25M raised total ($15M seed + $10M follow-on). Spark Capital and General Catalyst leading. Angels include the Collison brothers (Stripe), Logan Kilpatrick (DeepMind), Joanne Jang (OpenAI), the Cognition founders, and Guillermo Rauch (Vercel). - 10x Subscriber Growth
Subscribers grew tenfold over the past two months. Exact numbers aren't disclosed, but Poke hit #1 on the Vercel AI Gateway leaderboard. - No Model Lock-in
"Meta AI is locked to Meta's model, ChatGPT to OpenAI's. We pick the best model for each task" — von Hagen. A structural advantage over Big Tech competitors.
The technical backbone is Linq, a messaging integration solution — infrastructure that lets AI assistants live inside messaging apps. It raised $20M in February 2026. WhatsApp is still limited because Meta blocked general chatbots last fall, though that's loosening as EU, Italian, and Brazilian regulators launch antitrust investigations.
How Is It Different from Conversational AI?
Here's a side-by-side comparison — the same task, handled two different ways.
| Task | ChatGPT / Claude | Poke |
|---|---|---|
| Email alerts | Open the web or app and ask every time | Set "notify me when family or my boss emails" once → automatic forever |
| Morning schedule | Start a new conversation every day | Automatically arrives as a text each morning |
| Interface | Dedicated app or website | iMessage, SMS, or Telegram you already use |
| Learning curve | Have to learn how to write prompts | Search a Recipe → one click to install |
| Model choice | Platform-locked | Auto-routing to the best model per task |
Recipes are the real core. They're pre-built automation bundles across health, productivity, finance, scheduling, travel, home, school, email, community, and developer tools. One click to install, standard auth flow to activate. The integration list is impressive — Gmail, Google Calendar, Outlook, Notion, Linear, Granola, Strava, Withings, Oura, Fitbit, Philips Hue, Sonos, PostHog, Webflow, Supabase, Vercel, Devin, Sentry, GitHub, and Cursor Cloud Agents.
When users share their own Recipes, they earn $0.10–$1 per subscriber. A creator economy model, transplanted into AI agents.
The pricing model is interesting too — it's negotiation-based. In beta, users negotiated directly with the AI agent and landed somewhere between $10–$30/month (Poke answered the negotiation itself). Now it's automatically set based on usage patterns: free if you're not tapping real-time data (per-email processing, live flight check-in), more expensive if you use it heavily. Inference is where the real costs are, so this structure makes sense.
Poke says it runs regular penetration tests and multiple permission restrictions, but TechCrunch noted it "did not conduct its own security audit." By default, the team can't see your token contents — but if you explicitly toggle the log/analytics sharing switch, some access becomes possible. This is a tool handling sensitive health, financial, and email data, so trust is something only time can verify.
Getting Started
The setup is genuinely simple. You're done in 30 seconds.
- Go to Poke.com
Click "Get Started." No app download needed — that's the key differentiator. Pick your channel: iMessage (iPhone), SMS, or Telegram. - Enter your phone number → Send your first message
Once you're verified, just talk to Poke like you'd text a friend. "What's on my schedule today?", "Only notify me about emails from family or my boss" — that kind of thing. - Browse the Recipe directory
Head to poke.com/explore to see automations other users have built. Auto-tracking health data, daily morning news summaries, medication reminders — install any of them with one click. - Make it part of your routine
Two or three Recipes is plenty to feel the difference. The lowest-friction way to get started: once or twice a day, write out something you genuinely want automated and send it to Poke. No prompt engineering needed.
One more thing — it's useful for developers too. PostHog, Webflow, Supabase, Vercel, Devin, Sentry, GitHub, and Cursor Cloud Agents integrations are all built in, so you can get build failure alerts or error monitoring delivered straight to iMessage. The main market is non-technical users, but there's a real use case for devs as well.
There's a bigger implication here. Until now, AI agents have been a "for experts" category. Poke has brought the barrier to entry close to zero. When von Hagen says he wants to build "a product a billion people use," that's not just a vision statement — email and SMS are already infrastructure running at billion-user scale. Just put agents on top, with no model lock-in required. Big Tech has been trying to do this by locking users into their own models and apps; Poke went around them through messaging infrastructure instead.
Deep Dive Resources
TechCrunch Exclusive — Poke Goes Public Primary source covering the founding story, $25M raise, $300M valuation, Recipe system, and negotiated pricing model. Includes a von Hagen interview. techcrunch.com
Poke Official Site Where to get started — no install required. Browse the Recipe directory at poke.com/explore to see user-built automations. poke.com
Linq — The Infrastructure That Lets AI Live in Messaging Apps The $20M-funded backend that makes Poke possible. The technical stack behind iMessage and WhatsApp integration. techcrunch.com
TechBuzz — Poke Review and OpenClaw Comparison Real-world review from a non-technical perspective. Screenshot-heavy walkthrough of installing and using Recipes. techbuzz.ai
