Back
TechCrunchLaunchTechCrunch2026-06-30

Acti Brings AI Agent Shortcuts to iOS and Android Keyboards

Acti is a new keyboard app for iOS and Android that embeds AI agents directly into the typing experience, letting users create custom natural-language shortcuts that work across any app. The startup is betting that the keyboard — not a standalone app — is the right surface for ambient AI assistance.

Original source

Acti launched today with a keyboard replacement for iOS and Android that embeds AI agents at the point of typing rather than requiring users to context-switch to a separate app. The core feature is a shortcut system: users define triggers in natural language, and the keyboard executes them inline — drafting replies, reformatting text, translating, or pulling in external data — across any application on the device.

The product targets a real friction point. Current AI assistants on mobile require leaving whatever app you're in, copying text, pasting a prompt, then copying the result back. Acti's keyboard layer collapses that into a single gesture. The company claims shortcuts work system-wide, which would make it one of the few AI tools that doesn't require app-by-app integration.

Custom shortcut creation uses natural language, meaning users describe what they want the shortcut to do rather than configuring it through menus or code. Whether the agent execution quality holds up across diverse real-world tasks — the kind of messy, context-dependent requests that break most demo-era agents — is the open question. Acti hasn't published pricing details or clarified which underlying models power the agents.

The keyboard-as-platform bet is not new — SwiftKey and Gboard have experimented with AI features for years — but neither has leaned into agent-style task execution with user-defined shortcuts. If Acti's cross-app execution actually works reliably, it represents a meaningfully different approach to mobile AI interaction rather than another assistant app competing for home screen real estate.

Panel Takes

The Skeptic

The Skeptic

Reality Check

The demo sounds clean, but 'works across apps' on iOS is a load-bearing claim that Apple's sandboxing model has historically made very hard to deliver on. The specific scenario where this breaks is obvious: any app with a custom text input, rich text editor, or non-standard keyboard implementation — which is most productivity apps people would actually want this for. My prediction for what kills it in 12 months: Apple ships a native iOS keyboard AI layer in the next OS update and Acti's entire wedge disappears, because they've built a feature, not a platform.

The PM

The PM

Product Strategy

The job-to-be-done is sharp: execute AI tasks without leaving the current app. That's a real, specific problem with a measurable cost in friction, and the keyboard surface is a genuinely clever answer to it. The risk is completeness — if the custom shortcut creation requires more than 60 seconds to produce a working result on first try, users will abandon it before they ever experience the value, because mobile users have near-zero tolerance for setup flows that don't immediately pay off. No published pricing and no model disclosure means the product isn't complete enough to evaluate switching costs yet.

The Futurist

The Futurist

Big Picture

The thesis here is falsifiable and interesting: in 2-3 years, the dominant AI interaction layer on mobile will be ambient and cross-app, not siloed inside assistant apps — and whoever owns the input layer owns the session context. Acti is riding the trend of AI moving from destination to infrastructure, and they're early enough that the keyboard surface isn't yet crowded. The second-order effect that matters: if this works, it shifts power from app developers to the keyboard layer, because the keyboard starts intercepting and augmenting communication that previously stayed inside each app's walled garden.

The Builder

The Builder

Developer Perspective

The primitive is a system-level text transform with user-defined triggers — that's actually composable and interesting if the execution layer is exposed. What I need to know before I care: is there an API or developer SDK for building custom agent actions, or is this a closed shortcut menu with natural language dressing on top? If it's the latter, this is just a fancy text expander with a GPT call, and I can replicate 80% of it with a Shortcuts automation and an OpenAI API key. No docs, no repo link, and no model disclosure in the launch means I can't evaluate the craft yet — and that's a problem.

Bookmarks

Loading bookmarks...

No bookmarks yet

Bookmark tools to save them for later