Holmes

Holmes is an autonomous desktop agent designed specifically for macOS that observes user activity to understand context and proactively execute complex tasks. It operates continuously in the background, reading screen pixels, active windows, clipboard contents, and cursor position to determine what the user is working on. Holmes is intended for Mac users seeking a context-aware assistant that adapts to their workflow and habits over time.

The tool offers a range of capabilities including content creation, coding assistance, research and insights, and productivity support. Holmes provides a command bar accessible via keyboard shortcut, allowing users to issue questions and slash commands such as /run for executing multi-app workflows, /search for deep searches across apps and files, and /remind for setting context-aware reminders. The agent features a context menu and a "clippy style" desktop assistant interface, aiming to integrate seamlessly into the user's daily tasks. Holmes learns user preferences and writing style over days and weeks, further personalizing its assistance.

Privacy is a core focus, with all inference and processing performed locally on the device using CoreML. Holmes does not upload raw content or send screen data to the cloud, ensuring that user data remains on the Mac. The application is built natively for macOS using Swift and SwiftUI, and integrates into the menu bar and system workflow.

Holmes is currently available in a free beta version, granting full access while the tool is refined. Planned pricing tiers include a Pro plan at $15 per month with increased usage limits and early access to new features, and a Max plan at $50 with additional usage and advanced model tuning, both of which are noted as coming soon. Users can join a waitlist to access the beta. Holmes positions itself within the class of autonomous agents for desktop environments, focusing on proactive, context-driven task execution for macOS users.