Agentic Keyboard Glossary
Quick definitions of terms used in the agentic keyboard, AI keyboard, and mobile AI agent space. This glossary is a companion to the concept guide and the apps directory — use it to clarify terminology as you read.
Agentic Keyboard
A mobile keyboard that executes tasks on behalf of the user — searching, booking, sharing, composing — rather than just improving text. It understands typed intent and takes action across apps from the keyboard interface.
Why it matters: Represents a shift from keyboards as text-input tools to keyboards as command surfaces for mobile AI.
AI Keyboard
A keyboard that uses artificial intelligence to improve typing quality — through grammar checking, tone detection, next-word predictions, voice dictation, and translation. It improves the text you produce but does not execute tasks.
Why it matters: The most mature category of smart keyboards. Tools like Grammarly, SwiftKey, and Gboard are AI keyboards, not agentic keyboards.
Keyboard AI Agent
Another term for an agentic keyboard. Combines the concepts of an AI agent (a system that acts autonomously) with the keyboard form factor. Also called an 'AI keyboard agent.'
Why it matters: Different search queries use different word orders — 'keyboard AI agent' is a reverse-phrasing of 'AI keyboard agent.' Both refer to the same concept.
Voice AI Keyboard
A keyboard that uses AI-powered voice recognition and dictation as its primary or a major input method. Unlike basic voice-to-text, voice AI keyboards format spoken words into clean, structured text with punctuation and paragraph breaks.
Why it matters: Voice input is 2-3x faster than typing for many users. Wispr Flow is the leading voice-first AI keyboard. Gboard and SwiftKey also include voice dictation.
Companion AI App
A separate mobile app that provides AI assistance — drafting, research, summarization — but is not a keyboard replacement. ChatGPT and Perplexity are companion AI apps. They require switching between apps rather than working inline while typing.
Why it matters: Many agentic-like workflows are possible today using companion apps alongside keyboards. But companion apps add app-switching friction that true agentic keyboards aim to eliminate.
Mobile AI Agent
A general-purpose AI assistant that runs on a mobile device, capable of understanding intent and taking actions — searching, summarizing, drafting, analyzing — through a standalone app or system integration. Not specifically a keyboard.
Why it matters: Mobile AI agents like ChatGPT and Perplexity show what agentic mobile AI can do today, but they are separate apps. The agentic keyboard concept brings this capability into the typing surface.
Full Access
An iOS permission for third-party keyboards that enables network access and shared container access for the keyboard extension. Without Full Access, the keyboard is sandboxed — it cannot communicate over the network or access shared data containers, so cloud AI features and cross-app integrations do not work. Full Access does not by itself grant contacts, calendar, microphone, or location access; those are separate app and system permissions that the keyboard must request independently.
Why it matters: Most AI features — cloud predictions, voice dictation, web search, and cross-app actions — require Full Access on iOS. Understanding what Full Access does (and does not) grant helps users make informed decisions: enabling Full Access allows network communication, but it does not automatically give the keyboard your contacts, calendar, location, or microphone — those require additional permission prompts. Always review each permission separately.
On-Device Processing
AI computation that happens locally on your phone, without sending data to external servers. Text predictions, autocorrect, and some grammar checking can run on-device. More private than cloud processing.
Why it matters: On-device processing keeps sensitive text local and works offline. The trade-off: on-device models may be less capable than cloud models for complex tasks like web search or multi-step reasoning.
Cloud Processing
AI computation that sends data to remote servers for processing. Required for advanced AI features like web search, booking integrations, and large-model text generation. Data leaves your device.
Why it matters: Almost all advanced AI keyboard features require cloud processing. The privacy implication is that your typed content may be transmitted to company servers. Reputable keyboards publish policies describing what is sent and why.
Cross-App Workflow
A task that spans multiple apps — for example, searching a restaurant while messaging a friend, then booking a table and sharing the confirmation. An agentic keyboard executes cross-app workflows from the typing surface without requiring manual app switching.
Why it matters: Reducing app-switching is the core productivity promise of agentic keyboards. Each cross-app workflow replaces 3-5 manual app switches.
Intent-Aware Suggestions
Keyboard suggestions that go beyond predicting the next word — they predict what you want to accomplish (search, book, share, draft) and offer actions to complete those goals. A key difference between AI keyboards (which predict words) and agentic keyboards (which predict intent).
Why it matters: Intent awareness is what distinguishes agentic keyboards from AI keyboards. The keyboard must understand not just what you are typing but what you are trying to do.
Keyboard Extension
A software component that allows a third-party keyboard to run on iOS or Android. Keyboard extensions provide the framework for custom keyboards to appear in place of the system keyboard and access limited system resources. They are the technical foundation for both AI keyboards and agentic keyboards.
Why it matters: The capabilities of AI and agentic keyboards are constrained by what the operating system allows through keyboard extensions. iOS extensions are more restrictive than Android's, which affects feature availability across platforms.