Three terminals, three agents, three different projects. To know where any of them stand, you cycle through the windows and type the same question into each one. The agents work in parallel; you are the serial bottleneck polling them for status.
Why this keeps happening.
An agent's progress lives inside its session — a scrolling transcript in one terminal window. Nothing rolls up. Running agents in parallel multiplies the problem: every extra agent is one more window to check, one more “how's it going?” to type, one more answer that scrolls away. And when a session compacts or closes, even that answer is gone — the status existed only in a conversation that no longer does.
The way of working.
Progress is reported, not asked for
As agents work, they log completed steps, decisions, and open questions to SharedContext over MCP — at the moment each thing happens, not as a summary you have to request at the end.
One inbox across every project
Whatever any agent changed lands in one inbox. You scan it once and see what moved across all of your projects — nothing depends on which terminal you last looked at.
Project views hold the current state
Each project's detail view shows the active scope, recent session activity, and decisions waiting for your confirmation. Status is something you read, not something you interrupt an agent to reconstruct.
Superself is built with parallel Claude Code sessions — different workstreams, sometimes different repos, running at the same time. Each session logs its steps and decisions to SharedContext as it goes; we check the inbox and project views to see where everything stands. “How's it going?” is a question we stopped typing.
Start here.
See how session activity becomes shared state in How Superself works, then connect the Agents you already run in parallel.