Atua

Converts natural language commands into executable macOS automation sequences using on-device language processing, eliminating cloud round-trips. The system parses user intent, maps it to available system APIs and application hooks, and generates task workflows that execute locally with full access to system resources. This approach maintains privacy while enabling context-aware automation without latency penalties from cloud inference.

Monitors active application context and automatically adapts automation behavior based on which app is in focus, window state, and application-specific data. Uses macOS Accessibility API to introspect UI hierarchies, extract semantic information from application windows, and trigger app-specific automation hooks. This enables workflows that understand application state and respond intelligently without explicit user configuration per app.

Monitors clipboard content and automatically triggers automation workflows based on clipboard data, or populates clipboard with automation results for downstream use. Supports clipboard history tracking, clipboard format conversion (text to structured data), and clipboard-based data passing between automation steps. Enables clipboard-centric workflows where data flows through the clipboard without explicit file or database operations.

Supports defining automation workflows in multiple natural languages (English, Spanish, French, German, etc.), with the on-device language model translating non-English task definitions to a canonical internal representation. Enables non-English speakers to define automations in their native language without requiring English proficiency. Language detection is automatic, and users can switch languages per workflow or globally.

Maintains version history of automation workflows with the ability to view, compare, and rollback to previous versions. Supports branching and merging of workflow definitions for collaborative development. Tracks changes with metadata (author, timestamp, change description) and enables reverting to known-good versions if automation changes cause issues. Integrates with optional cloud sync for distributed version control.

Enables definition of multi-step automation workflows with branching logic, loops, and state-based decision points. Users can compose sequences of actions (application interactions, system commands, data transformations) with conditional branches based on task results, system state, or extracted data. The execution engine maintains state across steps and supports error handling and retry logic without requiring programming knowledge.

Directly invokes macOS system APIs and frameworks (Foundation, AppKit, Quartz) to automate system-level operations including file management, process control, system preferences, and inter-application communication. Bypasses the need for AppleScript or shell scripting by providing high-level abstractions over native APIs, enabling faster execution and deeper system integration than script-based approaches.

Specializes in automating repetitive research workflows including web scraping, data extraction from multiple sources, and structured data collection. Integrates with browsers and research tools to automate information gathering, deduplication, and organization into structured formats. Maintains research context across sessions and supports batch processing of research queries without manual intervention.

Provides a repository of pre-built automation templates for common tasks (email management, file organization, data entry) that users can customize and deploy. Templates are defined in a portable format that captures task sequences, conditional logic, and variable bindings. Users can share templates within teams or communities, and the system handles template versioning and compatibility across Atua versions.

Executes all automation logic and data processing on-device by default, with optional cloud synchronization for workflow definitions and execution logs. User data never leaves the device unless explicitly configured for cloud backup. The architecture maintains a clear separation between local execution (always private) and optional cloud features (workflow sync, cross-device access), giving users granular control over what data is shared.

Enables users to define automation tasks using natural spoken language, leveraging macOS speech recognition to convert voice input to text, then processing through the natural language task automation engine. Supports voice-based task triggering and status queries, making automation accessible without keyboard interaction. Voice commands are processed on-device using native macOS speech recognition APIs.

Allows users to bind automation workflows to custom keyboard shortcuts and system-wide hotkeys, enabling rapid task triggering without navigating menus or interfaces. Supports modifier key combinations and context-aware hotkey behavior that changes based on active application or system state. Hotkey bindings are managed through a centralized registry that prevents conflicts and enables easy remapping.

Enables scheduling of automation workflows to run at specified times, intervals, or in response to system events (file changes, network state changes, application launches). Maintains a background daemon that monitors triggers and executes scheduled tasks without user interaction. Supports cron-like scheduling syntax and event-based triggers, with execution logs and error notifications.