Poke

Poke is a personal AI agent that lives inside your text messages, built by The Interaction Company of California (Palo Alto). On June 4, 2026, Poke became the first AI agent approved by Apple for the Messages for Business platform, giving it native access to iMessage alongside SMS, Telegram, and WhatsApp (limited markets). The product launched publicly in March 2026 and has relayed over 100 million messages, with users growing 10x in recent months. The core idea is simple: text Poke like you would text a friend, and it handles tasks autonomously. Poke manages your calendar (Google Calendar, Outlook), drafts and sends emails (Gmail, Outlook), controls smart home devices (Philips Hue, Sonos), edits photos, checks you in for flights, tracks flight deals, sets reminders, summarizes YouTube videos, generates QR codes, and searches the web. It connects to Notion for notes and databases, Oura for health and sleep data, Strava for fitness tracking, Navan for travel management, GitHub for developer workflows, and Granola for meeting action items. The agent is proactive: it sends morning briefings, flags urgent emails, and triggers automations on schedules you define without waiting for a prompt. Poke is model-agnostic, selecting the AI model that best fits each task from providers including OpenAI, Anthropic, Google, and open source models. Co-founder Marvin von Hagen has stated this is a long-term competitive advantage over Meta AI (locked to Meta models), ChatGPT (locked to OpenAI), and Gemini (locked to Google). Poke uses Linq, a solution that enables AI assistants to live within messaging apps, as the underlying infrastructure for iMessage and other platform integrations. The Poke Recipes and Kitchen system enables extensible integrations. Developers can build custom MCP servers that connect into Poke, adding new tools and data sources that the agent can discover and call during conversations. An official MCP server template is published on GitHub (InteractionCo/mcp-server-template) with examples for no-auth, API key, and OAuth implementations. Users share recipes in a community marketplace covering categories from productivity and health to developer tools, finance, and travel. This makes Poke's integration surface open-ended rather than limited to a fixed list. Pricing is fully public. Free at $0/month forever includes connecting all apps, email, and calendar with no credit card required. Pro at $19/month adds frontier models for complex tasks, real-time automations running in the background, higher rate limits, and priority support. Ultra at $199/month provides frontier models powering every action, the highest rate limits, highest-priority support, and pay-as-you-go usage beyond included credits. Monthly and yearly billing are both available. Poke is not suited for teams needing enterprise-grade governance, compliance controls, or centralized admin management. Microsoft Scout (M365 E7 at $99/user/month) provides governed Entra identity, audit logging, and enterprise security controls that Poke does not offer. Gemini Spark (Google AI Ultra at approximately $100/month) provides deeper Google Workspace integration for teams already in that ecosystem. For teams needing a shared team inbox or customer-facing support agent rather than a personal assistant, Respond.io (from $79/month) or Intercom Fin ($0.99/resolution) serve different use cases entirely. Current state Q2 2026: Poke became the first AI agent approved for Apple Messages for Business on June 4, 2026. The service is available worldwide via iMessage, SMS, Telegram, and WhatsApp (limited markets due to Meta restrictions on third-party AI agents). Investors include Spark Capital and General Catalyst, with notable angels including John and Patrick Collison (Stripe founders), Logan Kilpatrick (DeepMind), Joanne Jang (OpenAI), and Scott Wu and Walden Yan (Cognition/Devin founders). Funding amount has not been publicly disclosed. No G2 or Capterra presence exists. No specific security certifications are published. Privacy policy confirms data may be used for AI model training, with a Maximum Privacy option that excludes user data from training.