Command a fleet of AI agentsfrom a single bridge.

00The problem

More agents should mean more done. Instead, the bottleneck is you.

• You can't see your own fleet. Six terminals, six tabs, six scrollbacks. Which one is done? Which one is stuck?
• The agent that needs you waits in silence. It asked for a permission twenty minutes ago, behind a window you weren't looking at.
• Your agents can't talk to each other. The frontend needs an endpoint from the backend, so you become the messenger.
• Work stops when you do. Close the laptop and nothing moves. The 8 am follow-up never gets sent.
• The results get buried. The report, the diagram, the plan you asked for scroll away under thousands of lines.

01See everything at a glance

Your whole fleet, on one map.

Every Claude Code session from every project in one window: what it's working on, how far along it is, how much context it has left and what it costs. Sessions group by repo, and the ones that need you rise to the top.

• Live status, branch and uncommitted changes
• Context left and cost, per session
• Who is handing work to whom, drawn as links

⌘E in the app

02Step in only when it matters

It taps you on the shoulder.

When an agent stops for a permission or a decision, you know the moment it happens, not twenty minutes later behind a tab you weren't watching. Click the notification and you're right there.

• Native macOS notifications
• Agents that need you jump to the top
• Cycle through everyone waiting on you

⌘⌥⇧↓ in the app

03Take the helm

The real Claude Code. Right in the app.

Open any agent and you're in its own terminal, inside OpenAdmiral. Not a wrapper, not a lookalike: the same Claude Code with your plugins, hooks, skills, MCP servers and shortcuts. Approve, answer, move on.

• Attach to background sessions instantly
• Every feature the day Claude Code ships it
• Prefer your own terminal? That works too

⌃1 in the app

04Agents that work together

Stop being the messenger.

An agent can hand a piece of work to another with a clear brief, and that agent reports back as it goes: accepted, in progress, blocked, done. The checkout needs an endpoint? It asks the API agent. You watch the work flow instead of carrying it.

• Delegations with status, notes and results
• See what's blocked, and on whom
• The sender hears back the moment work lands

⌘D in the app

05Work that keeps moving

Close the laptop. The fleet keeps sailing.

Line up what each agent should do next. A prompt runs when the agent is free, once another agent's work lands, or on a schedule, waking a closed session up if it has to.

• Queue prompts and slash commands
• Run once a delegation is done
• Recurring schedules: weekdays at 08:00

⌘J in the app

06Read the log

Every conversation, long after it scrolled away.

Each session's whole conversation, every prompt, tool call and reply, readable and searchable, even for the sessions you never attached to. Catch up in seconds on what an agent did while you were away.

• Turn-by-turn index with tool calls
• Search prompts and replies
• Works for sessions running anywhere

⌃2 in the app

07Every result, in one place

Rendered, not printed.

Reports, plans, diagrams and pages your agents produce are collected beside the work that made them, every version kept. No more hunting for the summary buried under ten thousand lines of output.

• Reports, UI mocks, diagrams and live HTML pages
• Every version kept, with notes
• One gallery for the whole fleet

⌘G in the app

08No surprises on the bill

Know what it costs before the bill does.

What each project costs today and this month, which models it runs on, and how much of your plan is left before the limit, always in view.

• Spend per project and per model
• Month-end projection
• 5-hour and weekly plan limits, with reset times

⌘U in the app