Mac companion
Rider
When several AI tools are working, keeping every window in view becomes a job of its own. Rider gives their activity a common place to arrive and a deliberate way to reach me.
- My role
- Creator & product designer
- The work
- Mac companion
The problem
Coding tools have different hooks, notifications, and ways of reporting progress. A completion, a request for attention, and a failed delivery should not all collapse into the same vague signal.
What I built
I built Rider as a local Mac runtime and interface around an EOX event hub. It receives activity from tools such as Claude Code, Codex, and Cursor, then routes useful updates to speech, notifications, or destinations I enable. Delivery state is part of the design: the system should distinguish an attempted message from one that actually arrived. Its connection to Erich OS also supports an approved path from a task to local work.
The result
Rider is the companion layer in my own working system. It reduces the need to watch every tool continuously while preserving control over what runs and where updates go. Raw prompts and transcripts stay local. The broader lesson is that coordination needs a feedback loop: a useful agent system must be able to tell its operator what happened.