Poke, a Palo Alto startup, runs AI agents entirely through text messages—iMessage, SMS, and Telegram—so users can set up automations without installing an app or managing API keys. The platform’s distinct technical choice is provider-agnostic model routing: it picks the model best suited to each task, including open-source models, rather than tying automations to a single vendor’s model.

How Poke decides which model handles a task

Poke implements a routing layer that examines a job’s requirements and dispatches it to the AI model that should perform best for that work. That can mean using a large commercial model for nuanced language tasks, a smaller open-source model for cheaper classification, or specialized providers for code edits—Poke’s architecture is explicitly provider-agnostic to avoid vendor lock-in.

This design matters because it creates a practical checkpoint: as new models improve or new vendors emerge, Poke can swap targets without forcing users to migrate their automations. The next thing to watch is whether Poke’s router keeps pace with fast-moving open-source releases and how it signals model choice to users when cost or privacy trade-offs change.

Using only text: recipes, connectors, and the non-technical user experience

Recommended Reading

Physical AI isn’t just “automation + ML”: modular pipelines are moving into production while end-to-end learning still needs more data and tougher hardware

Physical AI—robots that perceive, reason, and act—is being framed as the next step for factories. That framing is correct in intent but misleading in practice: production deployments today look nothing like simple add-ons to traditional automation, and the split between modular AI pipelines and holistic end-to-end learning models matters for who can adopt what, and […]

Physical AI isn’t just “automation + ML”: modular pipelines are moving into production while end-to-end learning still needs more data and tougher hardware

Users trigger and configure automations—called “recipes”—by text. Recipes install with a click and plug into services such as Gmail, Google Calendar, Outlook, Notion, Strava, Fitbit, Philips Hue, plus developer tools like GitHub and Vercel. The idea is to meet people where they already communicate: no new app, no dashboard, just messages.

That simplicity has practical trade-offs. It lowers onboarding friction and supports a creator economy (authors earn revenue from new sign-ups), but it also makes failure modes harder to debug because there’s no visual workflow builder. Poke is positioning this for everyday users and influencer-driven growth rather than as a developer-only tool like Zapier clones or SDK-first platforms.

Which actions cost money (and when you’ll likely keep using it for free)

Poke charges mainly for higher-infrastructure, real-time inference tasks—examples include live email monitoring, continuous flight check-ins, or ongoing sensor polling—while basic queries and non-live automations are often free. During beta, the company even let some users negotiate pricing with the system, a sign of prioritizing growth over immediate revenue capture.

Security, platform limits, and regulatory checkpoints

Poke uses layered controls—permission gates, limited-token access, and penetration testing—to restrict what the platform can do without explicit user opt-in. However, the company has not published an independent security audit, which is a meaningful gap for enterprise buyers who demand third-party verification before granting broad data access.

Operationally, Meta’s current restrictions on general-purpose bots limit WhatsApp support; Poke’s launch footprint therefore leans on iMessage and SMS in the U.S. Regulatory action in jurisdictions such as the EU, Italy, and Brazil could change WhatsApp’s availability and materially affect Poke’s international reach. Separately, Poke’s $300 million post-money valuation and backers—Spark Capital and General Catalyst—mean it has runway to expand, but larger-scale adoption will hinge on how it handles auditability, model-selection transparency, and compliance with messaging platforms’ changing rules.

Short Q&A

Can businesses rely on Poke for sensitive workflows? Not without additional assurances: Poke offers opt-in controls and penetration testing, but no public independent audit yet—enterprises should treat the platform as early-stage in security maturity.

Will WhatsApp work in the near term? Not fully; Meta’s restrictions limit general-purpose bots today. Watch regulatory moves in the EU, Italy, and Brazil for changes that could reopen WhatsApp access.

How is Poke different from other agent tools? The combination that matters is a text-only interface plus provider-agnostic model routing and turn-key recipes aimed at non-technical users—this mix targets mainstream, no-install automation rather than developer-first or single-vendor agent ecosystems.