SKYLINE
fleet control for Claude Code · by High Country Codes
If you run one Claude Code terminal, you don't need Skyline. If you run six of them across six repos and you're alt-tabbing to remember which agent is doing what, Skyline is the missing layer: one place that sees the whole range.
What it does
- A durable job queue — dispatch work to headless runners that survive crashes, resume their own sessions, and report results with per-job cost receipts.
- Model-lane budgeting — one model's drained window parks only that lane; mechanical work routes to cheap models overnight.
- A needs-decision inbox and a human-task list — your agents file the things only you can do; your phone buzzes; your one-line answer unblocks them.
- An agent on your machine — Windows (WSL), macOS, Linux. Your repository files and git history are never uploaded, and your Claude login stays local. What the dashboard shows is the activity your agent streams out — repo and session names, job and cost status, activity notes, the step feed (which includes the file paths, commands, and code fragments it works with), and the chat you send and its replies — relayed through Skyline Cloud and stored there to render your dashboard, kept until newer activity replaces it.
Skyline drives the Claude Code CLI you already have, under your own login; it makes no Anthropic API calls and holds no keys of its own. Your repository files and git history are never uploaded. To render your dashboard, Skyline Cloud relays and stores the activity you watch — repo and session names, job and cost status, activity notes, the step feed (file paths, commands, and code fragments included), and chat and its replies — kept only to display your fleet, until newer activity replaces it. Repos you mark confidential never leave your machine. We run no product analytics and sell nothing to anyone.
Private beta. Skyline is an independent product of High Country Codes and is not affiliated with or endorsed by Anthropic.