Your AI-powered dev fleet Head Quarters.
Track all your projects, every Copilot session, every tool call — from one dashboard.
Today it's about visibility, productivity and peace of mind. Tomorrow, we'll add Context Hub — where your context, skills and tools follow every agent safely, everywhere.
npx launchpad-hqOr install globally:
npm install -g launchpad-hq
launchpad-hqRequirements: Node.js 18+ and GitHub CLI authenticated (
gh auth login)
Opens at http://localhost:4321 — a three-pane mission control dashboard. Add your GitHub repos, see their issues on a kanban board, and watch Copilot sessions live as they work.
Browser ←ws→ HQ Server ←ws→ Daemon(s)
- HQ Server — Fastify backend serving a React dashboard. Aggregates state from all projects. Talks to GitHub's GraphQL API for issues/PRs.
- Daemon (per project) — Lightweight process running inside each project's devcontainer. Discovers Copilot sessions via
@github/copilot-sdk, streams events, relays terminal I/O. Connects outbound to HQ over WebSocket. - WebSocket Protocol — Dual-socket:
/wsfor browser clients (channel-based pub/sub),/ws/daemonfor daemon connections (auth handshake + typed protocol). - React UI — Mantine + TanStack Router/Query. Three-pane layout: project list → kanban board → live sessions. Light/dark theme.
- Multi-project dashboard — All your repos in one view with attention badges
- Kanban boards — GitHub Issues auto-classified into Todo / In Progress / Done
- Live Copilot session introspection — See conversations, tool calls, agent activity in real-time
- Session steering — Inject prompts or attach to a terminal and take the wheel
- Daemon architecture — Hub-and-spoke: one HQ, one daemon per project environment
- Remote access — Dev Tunnels integration with QR code for phone/tablet access
- Terminal relay — Full xterm.js terminal takeover via the daemon
- Settings UI — Configure state mode, Copilot preferences, models, and tunnels
- Onboarding wizard — Interactive CLI setup on first run
- App preview proxy — See running apps from the dashboard
We're building toward a future where HQ doesn't just observe your AI fleet — it becomes your Context Hub.
- 🧠 Context Hub — MCP servers, custom instructions, shared skills — configured once in HQ, deployed to every agent automatically (#59)
- 📊 Token usage & observability — Know where your AI spend goes across all projects (#60)
For the full vision, see VISION.md.
We develop this project openly using .squad/ — our AI team's memory lives in the repo. PRs welcome.
This is a build-in-public project. Arjen is building the future of AI dev tooling with AI itself. If that sounds interesting, jump in.
This repo uses Squad for AI-assisted development. The .squad/ directory contains team configuration, agent histories, and architectural decisions — it's how our AI team maintains context across sessions.
See .squad/team.md for the team roster.
MIT