Poke raised $25 million at a $300 million valuation. The story everyone is covering is the consumer app. The story nobody is covering is the plumbing underneath it.
Poke, the AI agent accessible via iMessage and SMS, added $10 million to its seed round in April, bringing total funding to $25 million on a post-money valuation the company says is $300 million. During the beta phase, more than 6,000 Silicon Valley insiders tested Poke, sending roughly 200,000 messages monthly, according to TechFundingNews. The company behind it, The Interaction Company of California, launched publicly in March and has attracted backing from Spark Capital and General Catalyst, plus angel investors including John and Patrick Collison (Stripe), Logan Kilpatrick from DeepMind, Joanne Jang from OpenAI, and Scott Wu and Walden Yan from Cognition. The team is 10 people. Co-founders Marvin von Hagen, 23, and Felix Schlegel, 25, previously built TUM Boring, a tunnel-boring project from Technical University of Munich that won Elon Musk's Not-a-Boring competition in 2021. Von Hagen has a track record in AI safety research: he is known for exposing the hidden "Sydney" persona inside Microsoft's Bing chatbot and raising early questions about prompt injection vulnerabilities in large language models.
But the more durable business may be the infrastructure layer beneath Poke.
Poke runs on Linq's messaging API, which connects AI agents to iMessage, SMS, and Telegram. Linq, based in Birmingham, Alabama, raised a $20 million Series A in February led by TQ Ventures and formally pivoted from B2B business messaging to AI agent infrastructure after Poke went viral on its API. The company's customers' AI agents now reach 134,000 monthly active users, process more than 30 million messages per month, and show net revenue retention of 295 percent with zero churn. Net revenue retention above 100 percent means existing customers are expanding their usage faster than new customers are being added. Zero churn means every customer Linq had six months ago is still a customer today.
These are the metrics of a company that has found product-market fit in the infrastructure layer of the AI agent stack. And they are almost never mentioned in coverage of the consumer app built on top of them.
The dependency raises a question the Poke funding story leaves unanswered: who captures the durable value in a consumer agent market that runs over someone else's API? Linq is the company that had to exist first, and that other AI agents now depend on to reach users without an app download.
The origin story is instructive. Linq launched as a digital business card for sales teams, pivoted to helping businesses send blue-bubble iMessages to customers instead of green-bubble SMS, and had no particular plans for AI agents. Then Poke's Interaction Company came along in spring 2025 looking for a way to reach iMessage users without building an app. When Poke went viral in September, the flood of inbound requests from other AI companies wanting the same iMessage access forced a decision: stay a B2B messaging tool or become infrastructure for a new category. Linq chose the latter.
Potter, Linq's CEO, has framed it as a question of position. "Do we stay a spoke of this wheel, or do we build the hub?" he told TechCrunch at the time.
The hub is still being stress-tested. Meta barred general-purpose chatbots from WhatsApp last fall, and Poke's access is currently limited. Regulators in the EU and Brazil have opened antitrust investigations into the decision, and Poke has returned to the Brazilian market as a result. Whether those cases reopen WhatsApp at commercially viable terms is an open question. The broader platform risk for Linq is obvious: the company that routes AI agents through Apple and Google's messaging platforms is subject to the same platform decisions that just blocked its biggest customer. Linq's stated ambition extends beyond messaging: programmatic voice, email, Slack, and Discord are all on the roadmap, but none of those channels currently generate the volume that iMessage does.
None of this is to say Poke is a bad bet. A 23-year-old founder who exposed an AI alter ego before most people knew what prompt injection was, building a consumer AI agent that people can use without installing anything, backed by the people who built Stripe and OpenAI, is not a speculative bet. The Collison brothers and the Cognition founders are not writing checks casually.
But the infrastructure story is cleaner. Linq's 295 percent net revenue retention and zero churn predate Poke's public launch and survived the WhatsApp block. The company that started as a digital business card is now processing 30 million messages a month for AI agents it had never heard of two years ago. That is the shape of a platform business, built accidentally, in the layer that the current wave of AI agent coverage is almost entirely ignoring.
Poke is the app. Linq is the pipe. And in infrastructure, the pipe is often the more durable asset.