Inspiration
Everyday work on macOS still involves unnecessary friction: files are dragged between windows, temporary items clutter the desktop, useful clipboard content gets lost, and simple actions require opening several separate applications.
CompanionKit was created to make these small workflows feel calmer and more direct. Instead of adding another traditional productivity dashboard, it introduces a lightweight desktop companion that stays accessible and helps users interact with files, clipboard content, and temporary items through a simple native interface.
What it does
CompanionKit is a native macOS desktop companion designed around three everyday workflows:
- Drop Zone: Drag files or content onto the companion to access relevant actions.
- Temporary Shelf: Hold files temporarily without moving or duplicating them unnecessarily.
- Clipboard History: Recover and reuse recently copied content.
The goal is to reduce context switching and make common desktop actions available from one coherent interaction point.
During OpenAI Build Week, CompanionKit is being developed with Codex and extended with GPT-5.6-powered contextual intelligence. The intelligent Drop Zone is designed to understand the type and content of an item and suggest useful actions instead of presenting the same generic menu for every file.
For example, the intended workflow can distinguish between a document, image, screenshot, or text item and offer actions appropriate to that context. The user remains in control and confirms actions before they are executed.
How we built it
CompanionKit is being built as a native macOS application using Swift and SwiftUI.
Codex has been used throughout the development process to:
- analyze and structure the application architecture,
- implement native macOS interactions,
- build and refine individual features,
- create tests and validation steps,
- identify defects and propose focused fixes,
- improve documentation and repository quality.
The product direction, scope, interaction model, and final implementation decisions remain human-directed. Codex serves as the primary engineering collaborator rather than replacing product judgment.
GPT-5.6 is used for the contextual reasoning layer of the intelligent Drop Zone, where unstructured content must be interpreted and converted into a small set of relevant, understandable actions.
Challenges we faced
One major challenge is balancing intelligence with simplicity. A desktop companion should reduce friction, not introduce another complex interface.
Native macOS development also requires careful handling of:
- drag-and-drop behavior,
- clipboard access,
- file permissions,
- application state,
- privacy,
- predictable execution of suggested actions.
Another challenge is ensuring that AI-generated suggestions remain transparent and controlled. CompanionKit is therefore designed around user confirmation rather than silently executing potentially destructive actions.
Accomplishments that we're proud of
- Creating a coherent native macOS product instead of a standalone chatbot.
- Combining file handling, temporary storage, and clipboard workflows in one interaction model.
- Building the project through a documented human–Codex development workflow.
- Designing the AI layer as contextual assistance rather than an intrusive replacement for user control.
- Establishing a foundation that can remain useful as a free and open-source macOS tool.
What we learned
We learned that effective AI-native software is not defined by the amount of AI it contains. The greater challenge is identifying the precise moment where reasoning adds value.
For CompanionKit, that moment occurs after the user provides an item but before they decide what to do with it. GPT-5.6 can interpret the context, while the native application provides a controlled and understandable way to act on that interpretation.
We also learned that Codex is most effective when autonomous implementation is combined with precise scopes, acceptance criteria, testing, and human product decisions.
What's next for CompanionKit
The immediate focus is to complete and validate the contextual Drop Zone, strengthen the overall native macOS experience, and prepare a reliable test build.
Future development may include:
- additional context-aware actions,
- configurable local and cloud processing,
- German and English localization,
- workflow extensions,
- stronger privacy controls,
- community-contributed actions and integrations.
The long-term objective is a small, calm, and useful desktop companion that removes friction without becoming another complicated productivity system.
Log in or sign up for Devpost to join the conversation.