Faros Desktop visualizes AI traces on macOS
Sachin spent the last couple of months building Faros Desktop a Mac application that visualizes how developers use local AI coding tools. The project emerged from a need to understand what belongs in session traces as agents replace hands-on-keyboard work. The app consolidates history from Codex Claude Code Github Copilot and Cursor into one interface.
Session traces from four AI coding tools
Faros Desktop hydrates with on-device session history letting users deep dive into a visualization of AI session traces. A user can jump around a session timeline and see exactly the messages and tool calls recorded in each trace. The organizer groups entries by project and task while on-device models summarize sessions in chunks. No API keys or login are required the app runs entirely locally.
Power Ups turn usage patterns into savings
Beyond visualization Faros Desktop surfaces patterns in usage and recommends simple changes the team calls Power Ups. These suggestions help users make their subscription spend go further through configuration adjustments that reduce unnecessary token consumption. The recommendations stem from consolidating the consolidated information across multiple coding sessions.
Seven years of Token Engineering Platform expertise applied to Builders
Faros has served large engineering organizations for the past seven years with a multi-player Token Engineering Platform. This Show HN product extends that expertise to individual Builders not just enterprise engineering orgs. The business model drives thought leadership and creates leads for the sales team rather than charging users for the desktop app itself.
On-device models summarize and organize sessions
The application uses on-device models to summarize local sessions in chunks then organizes them by project and task. This approach keeps data on the users machine while still providing meaningful overviews of AI-assisted development workflows. The app also only uses on-device models to summarize local sessions in chunks then organizes them by project and task.
Future directions: memory portability and team sharing
One direction the team is exploring involves using previous sessions to make the next one better. They are looking at local memory portability across AI coding toolchains as well as multi-player modes to share memory within a direct team. The goal is to let earlier sessions inform and improve later ones without sending data off the device.