Summary Cotabby is an open-source, on-device AI autocomplete application for macOS, providing system-wide inline text completion. It targets users seeking enhanced typing productivity and privacy by running all AI inference locally, eliminating cloud dependencies and ensuring data stays on the user's Mac.

How It Works This menu bar app renders AI-generated text continuations as "ghost text" next to the user's caret, accepting suggestions via Tab. It offers two inference engines: Apple Intelligence (macOS 26+) utilizing FoundationModels, and an Open Source engine running local GGUF models via llama.cpp (through llama.swift). Visual context is incorporated using screenshot OCR for more relevant suggestions, all optimized for low latency on Apple Silicon.

Quick Start & Requirements Install by downloading the Cotabby.dmg from GitHub Releases and dragging the app to Applications. Grant Accessibility, Input Monitoring, and Screen Recording permissions. Requires macOS 15.0 or later (macOS 26+ for Apple Intelligence engine). Apple Silicon is recommended for local model performance.

Highlighted Details

• System-wide inline autocomplete in any macOS text field.
• "Ghost text" UI displays suggestions as translucent overlays.
• 100% local inference ensures data privacy.
• Visual context via screenshot OCR enhances suggestion relevance.
• Supports built-in and custom GGUF models.

Maintenance & Community Developed by University of Waterloo students (@FuJacob, @jam-cai), Cotabby is currently in beta. While a Slack community is mentioned, specific links are absent.

Licensing & Compatibility Licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). The AGPL's network-use clause requires modified versions distributed over a network to also be source-available under AGPL-3.0, potentially restricting commercial integration.

Limitations & Caveats The application is in beta, requires specific macOS versions (15.0+, 26+ for Apple Intelligence), and necessitates sensitive system permissions. The AGPL-3.0 license imposes significant obligations for network distribution of modified versions.