A new startup called Poke is making AI agents accessible to everyday people by letting them control a personal assistant entirely through text messages. Instead of installing software or managing complex systems, users simply text Poke to automate tasks like calendar management, smart home control, medication reminders, and email filtering. The approach sidesteps the technical barriers that have kept agentic AI (AI systems that take independent action on your behalf) confined to developers and tech-savvy users .

Poke, backed by Spark Capital and General Catalyst, just raised $10 million in new funding on top of a $15 million seed round from last year, bringing its valuation to $300 million post-money. The 10-person startup launched publicly in March 2026 and has already attracted thousands of users creating custom automations through the platform .

Why Is Poke Different From Other AI Agent Frameworks?

The agentic AI space has exploded with tools like OpenClaw and LangChain, but these systems typically require developers to write code, manage dependencies, and troubleshoot errors through a terminal. They also raise security concerns because they often need deep access to your computer's operating system. For most people, this complexity makes agentic AI feel out of reach .

Poke eliminates these friction points by operating entirely through messaging apps. There's no app to download, no terminal commands, and no technical setup. Users visit Poke.com, enter their phone number, and start texting requests. The platform works across iMessage, SMS, and Telegram, with WhatsApp support available in some regions .

"What we noticed there was that people wanted to use Poke for everything. Even though it was only meant for email, people started asking Poke to remind them to take their medication. They asked Poke about sports results. And at that time, we didn't have a lot of this functionality, but we noticed how we needed to become general-purpose much more quickly, because people just like the personality and the humanness of it so much," said Marvin von Hagen, co-founder of The Interaction Company of California.

Marvin von Hagen, Co-founder of The Interaction Company of California

The startup's origin story reveals how user behavior shaped the product. Von Hagen and his team initially built Poke as an AI assistant for email management. But beta testers began asking it to do everything else, from medication reminders to sports updates to weather-based outfit suggestions. Rather than restrict the tool, the team pivoted to make Poke a general-purpose agent .

How Does Poke Actually Work Under the Hood?

Poke uses a flexible approach to AI models that sets it apart from competitors. Rather than locking users into a single provider's models, Poke selects the AI model that best fits each task, whether that's a model from OpenAI, Anthropic, Meta, or an open-source alternative. This model-agnostic strategy gives Poke a competitive advantage over platforms like ChatGPT, which only use OpenAI models, or Meta AI, which only uses Meta models .

The platform leverages Linq, a solution that enables AI assistants to function within messaging apps, allowing Poke to operate seamlessly across different messaging platforms. Users interact with Poke through pre-built "recipes" (automation templates) or by writing their own automations in plain text .

What Can Poke Actually Do for Users?

Poke's capabilities span multiple life and work domains through its recipe system. At launch, the platform offers pre-made automations across several categories:

• Health and Wellness: Track fitness goals, send medication reminders, and integrate with devices from Strava, Withings, Oura, and Fitbit
• Productivity and Work: Manage calendars, filter emails, and automate workflows with integrations to Gmail, Google Calendar, Outlook, Notion, Linear, and developer tools like GitHub and Vercel
• Smart Home Control: Operate connected devices from Philips Hue, Sonos, and other smart home manufacturers
• Finance and Travel: Monitor spending, track flights, and manage bookings
• Photo Editing and Content: Edit images and manage media files
• Developer Tools: Automate workflows with PostHog, Webflow, Supabase, Devin, Sentry, and Cursor Cloud Agents

Users can also create custom automations by writing instructions in plain text. The platform encourages community contribution by paying creators between 10 cents and a dollar per user signup for recipes they share .

How Does Poke Handle Security and Privacy?

Security is a critical concern for any system that takes autonomous action on your behalf. Poke implements a multi-layered security model that includes regular penetration testing, security checks, and permission limits for both AI agents and human employees. By default, Poke cannot see the contents of user data or tokens unless users manually opt in to share logs or analytics through their settings .

This privacy-first approach contrasts with some agentic systems that require broad system access. However, it's worth noting that TechCrunch has not performed an independent security audit of Poke's infrastructure .

What Does Poke Cost, and Who Is It For?

Pricing for Poke is surprisingly flexible and affordable. The service is free to start, with costs based on how the agent is being used. Tasks that don't require real-time data processing are free, while features that demand real-time inference, like automations triggered by every incoming email or live flight tracking, incur costs. During beta testing, users negotiated pricing directly with the AI agent, with monthly costs ranging from $10 to $30 .

Von Hagen emphasized that profitability is not the immediate goal. "We really don't want to make money, but we really want to grow. We want to build a product for a billion people and monetization is really secondary," he stated. The company's focus for the coming months is bringing Poke into everyday life through partnerships with creators and influencers .

Von Hagen

Why Does Poke Matter for the Future of AI Agents?

Poke arrives at a critical moment for agentic AI. Demand for these systems is surging, with OpenAI acquiring OpenClaw's creator and Nvidia CEO Jensen Huang declaring that every company needs its own agentic AI strategy. Yet most agentic systems remain inaccessible to non-technical users .

By making agentic AI accessible through a familiar interface, Poke could accelerate mainstream adoption of autonomous AI systems. The platform demonstrates that the barrier to agentic AI adoption isn't technical capability, but rather user experience and accessibility. If Poke succeeds in bringing agentic AI to a billion users, it could reshape how people think about automation and personal productivity.

The startup's success will also depend on navigating regulatory challenges. Meta's restrictions on general-purpose chatbots in WhatsApp have limited Poke's reach in some markets, though antitrust probes from the European Union, Italy, and Brazil may eventually force Meta to reconsider those policies. For now, Poke is available in most regions through iMessage, SMS, and Telegram, with WhatsApp access expanding as regulatory pressure mounts .