Mission control for your AI sessions.
Is it building what you described — or has it been drifting in the wrong direction for the past forty minutes? Sessions, campaigns, plans — one screen that shows you what's happening, in plain view. The panel below is a staged, fictional demo. Click through it; the real one renders your own sessions, live on your machine.
▸aurora-crmYou asked for a feature, walked away, and made coffee. Is it building what you described — or has it been looping on the wrong thing for forty minutes, quietly burning your plan? With everything in a scrollback you can't read, you find out when it's done. Or when a customer does. The panel answers “is this actually going the way I think it is?” — the session ticking through your request, the acceptance gate holding or looping back, in plain view. You don't read the terminal. You read the room.
You shouldn't need a terminal to know what's happening.
The tooling does the work either way. The panel is for watching it — and what you need to see depends on where you're coming from.
Claude says “take a look at the blueprint.” Which file? Where?
It's already in the panel — rendered, readable, organized by plan. No Finder expedition, no asking Claude to repeat the path.
Is it actually working, or did it stall ten minutes ago?
Watch the session itself: tasks ticking off, the campaign moving phase to phase, the acceptance gate holding or looping back. No terminal archaeology.
What changed while you were away?
The board answers at a glance — what moved into progress, what passed review, what's still queued. Read state, not scrollback. When the enforcement layer catches something — a leaked key, a cross-tenant hole — the panel is where you see it. See what's been caught →
Claude spawned three agents. Doing what, exactly?
The fleet view shows who's driving and what each delegate is on — the same picture Claude has, without learning to read a terminal to get it.
The free tooling runs all of this in the terminal — graph, blueprints, campaigns. The panel is the operational layer on top: the pane of glass that makes the work watchable. That layer is what Pro is.
What each tab shows you
Sessions
Every live Claude session and its agent fleet — who's driving, who's delegated, what each subagent is doing right now, connected the way the work actually flows.
Graph
Your codebase as a living map — files clustered by domain, dependency direction visible, the same graph the agents query before they touch anything.
Plans
Blueprints, tasks, audits, and campaigns on one board — what's planned, what's in progress, what passed review, straight from the state the agents work from.
Memory
What the work has learned and decided, kept as a map instead of lost in old chats — the decisions made, the patterns found, the things still to do. Your AI reads from it so the tenth session remembers what the first one figured out.
Campaigns
Bigger jobs that run themselves through stages — building, checking, looping back if a stage doesn't pass. Each card shows where its job is: running, paused, failed and why, or done.
Plugins
Everything switched on for this project — the Composure suites doing the work, and the guardrails (hooks) running on every change. What's installed, at a glance.
Account
Your plan and what you've used this month — reviews run, sessions watched — across the machines you work on. No surprises at renewal.
It runs on your machine. Nothing is transmitted. The panel reads the same local files your agents already read — sessions, the graph, your plans. No external calls, no telemetry, no copy of your code leaving your laptop. Watching the work doesn't mean uploading it.
The real one comes with Pro.
Everything that runs the work is free. Pro is the pane of glass that keeps you in the room.
New to building with AI? Start here →