What I built
A shared planning layer for AI work sessions: projects, spaces, docs, tickets, dependencies, notes, logs, and status.
It is local-first and designed so multiple AI tools can work from the same backlog.
Back
Problem
Agents can plan, but without shared durable task state they duplicate work, lose dependencies, and need the user to manually coordinate the queue.
Story
How it works
Create a project, break work into tickets, assign model tiers, claim the next ticket, log progress, and inspect state from the CLI or dashboard.
What it proves
The output is an explicit, claimable work queue with enough structure for multiple agents to avoid stepping on each other.