The starting point
Working with several AI assistants means each one starts a session without context, and a decision made in one chat is lost to the others. The goal was one local place that holds each project's context and that every assistant can read and update.
What we were responsible for
We designed and built the app, its MCP bridge and the migration from the earlier version.
What we built
-
Project library
Searchable projects with status, priority, next step and full history, each with an overview, history and analytics.
-
MCP bridge
A bundled server that any MCP client can connect to. Assistants read a project before work and save the outcome, the reasoning and next steps after it.
-
Safe writes
Each write appends history and updates the overview in one transaction, checks revisions and ignores exact retries; damaged data shows an error instead of being reset.
-
Rebuilt as a native app
An earlier Tauri and React version was rebuilt in Swift, with a verified migration and full backups of the old library.
Where it stands
Runs locally on macOS with native tests on synthetic data, and exports every project and its full history.