Drag a task → your agents execute → the board reflects reality.
Get Started · Features · Demo · Roadmap
You're running a company with AI agents — research, engineering, deals, treasury. But right now you're juggling:
| Without Shipyard | With Shipyard |
|---|---|
| Scattered agent outputs across Telegram, Slack, email | One dashboard, all agents, all activity |
| No idea what your agents are working on | Kanban board that updates in real time |
| Manually kicking off agent tasks | Drag to "In Progress" → agent activates |
| No cost visibility until the bill arrives | Per-agent budgets with approval gates |
| Copy-pasting context between tools | Agents share memory, goals, and workspace |
Drag a task to "In Progress" and your agent picks it up. When it's done, the task moves to "Done" automatically. Budget checks and approval gates built in.
Five specialized agents out of the box — Chief of Staff, Scout (research), Builder (engineering), Deal Flow (partnerships), Baron (treasury). Add your own with custom roles, skills, and accent colors.
Morning briefing in one glance: agent status (live pulsing dots), task summary, project health, recent intel, inbox, and goals — all on one screen.
Set monthly spend limits per agent. Shipyard checks budgets before activation and warns you when agents approach their limits. No surprise bills.
CI status, open PRs, and deploy health for your pinned repos. Color-coded badges update automatically.
Market intelligence reports from your Scout agent. Deploy research runs from the dashboard.
Google Calendar events and important emails surfaced automatically. No more tab-switching.
Browse your agents' memory files, daily notes, and long-term context. Everything is transparent.
Talk directly to your agents from the dashboard. Ask questions, give instructions, get status updates.
Responsive design with bottom navigation. Manage your agent team from your phone.
npx create-shipyard-appThis will:
- Clone the repo
- Install dependencies
- Launch the setup wizard
- Open your browser
git clone https://github.com/BiscayneDev/shipyard-os.git
cd shipyard-os
npm install
npm run devOpen localhost:3000 — the setup wizard walks you through connecting your agents.
git clone https://github.com/BiscayneDev/shipyard-os.git
cd shipyard-os
npm install
npm run devIn the setup wizard, choose "Ollama / Local Models" and point it at your running Ollama instance. Works with LM Studio and llama.cpp too — anything that speaks the Ollama API.
Don't have agents set up yet? No problem. Click "Skip — explore without agents" in the setup wizard to get a feel for the dashboard with sample data.
┌─────────────────────────────────────────────────────────┐
│ SHIPYARD OS │
│ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Backlog │───▶│In Progress│───▶│ Done │ │
│ │ │ │ │ │ │ │
│ │ • Research│ │ 🔭 Scout │ │ ✅ Report │ │
│ │ • Build │ │ running │ │ ready │ │
│ │ • Deploy │ │ │ │ │ │
│ └──────────┘ └────┬─────┘ └──────────┘ │
│ │ │
│ ┌────────▼────────┐ │
│ │ Budget Check │ │
│ │ Approval Gate │ │
│ └────────┬────────┘ │
│ │ │
│ ┌────────▼────────┐ │
│ │ Agent Runtime │ │
│ │ OpenClaw/Ollama │ │
│ └─────────────────┘ │
└─────────────────────────────────────────────────────────┘
- You drag a task to "In Progress"
- Budget check runs — does this agent have spend remaining?
- Approval gate shows you the cost estimate — confirm or cancel
- Agent activates via OpenClaw gateway
- Task auto-moves to "Done" when the agent finishes
- Toast notification fires — activity logged to the feed
No polling. No manual follow-up. The board is the source of truth.
| Layer | Technology |
|---|---|
| Framework | Next.js 16 (App Router) |
| Language | TypeScript (strict mode) |
| Styling | Tailwind CSS 4 |
| Drag & Drop | @hello-pangea/dnd |
| Agent Runtime | OpenClaw, Ollama, or custom |
| Storage | Vercel KV (prod) / local JSON (dev) |
| AI SDK | Vercel AI SDK + Anthropic |
| Icons | Lucide React |
| Deploy | Vercel |
| Agent | Role | Color |
|---|---|---|
| 🦞 Vic (Chief of Staff) | Orchestrates the team, delegates tasks | Purple |
| 🔭 Scout | Market intelligence, research, signals | Cyan |
| ⚡ Builder | Engineering, shipping code, CI/CD | Emerald |
| 🤝 Deal Flow | Partnerships, opportunities, outreach | Amber |
| 🏦 Baron | Treasury, wallet ops, financial tracking | Pink |
Agents are fully customizable — change names, roles, skills, and colors from the Agents page.
shipyard-os/
├── app/
│ ├── dashboard/ # Command center overview
│ ├── tasks/ # Kanban board with drag-to-activate
│ ├── agents/ # Agent team management
│ ├── projects/ # GitHub repos + CI status
│ ├── intel/ # Scout research reports
│ ├── calendar/ # Google Calendar integration
│ ├── chat/ # Direct agent chat
│ ├── settings/ # Config + reset wizard
│ ├── setup/ # First-run onboarding wizard
│ └── api/ # 15+ API routes
├── components/
│ └── Sidebar.tsx # Desktop sidebar + mobile bottom nav
├── lib/
│ └── config.ts # Runtime configuration
├── data/
│ ├── tasks.json # Local task store
│ ├── agents.json # Agent definitions
│ └── goals.json # Goal tracker
└── public/
- Kanban drag-to-activate with budget gates
- Agent team management (CRUD)
- Setup wizard with demo mode
- Post-setup health checks
- Budget enforcement + approval modals
- Gmail inbox integration
- Google Calendar integration
- Agent runtime abstraction — plug in OpenClaw, Ollama, or custom runtimes
- Agent marketplace — share and install agent identities
-
Cmd+Kcommand palette — quick actions from anywhere - Webhooks — real-time updates instead of polling
- Docker image —
docker run -p 3000:3000 shipyard-os - Multi-workspace — manage multiple agent teams
- Mobile app — native iOS/Android companion
We'd love your help making Shipyard OS better. See CONTRIBUTING.md for guidelines.
# Fork the repo, then:
git clone https://github.com/YOUR_USERNAME/shipyard-os.git
cd shipyard-os
npm install
npm run devDo I need OpenClaw to use this?
No! Shipyard OS supports multiple runtimes: OpenClaw for full agent orchestration, Ollama/LM Studio/llama.cpp for local models, or demo mode to explore without any runtime.
Is this just a dashboard?
No — it's an operating system. When you drag a task to "In Progress", it actually fires an agent. Budget gates prevent runaway costs. The board reflects what your agents are doing in real time.
Can I use my own agents instead of the defaults?
Yes. Add, edit, or delete agents from the Agents page. Each agent is a markdown identity file in your workspace — fully customizable roles, skills, and personalities.
What about costs?
Shipyard OS tracks per-agent token usage via the Anthropic Admin API. Set monthly budgets per agent, and the system will block activation when limits are reached. No surprise bills.
Can I self-host this?
Yes. It's a standard Next.js app. Run it locally, deploy to Vercel, or host anywhere that supports Node.js. Task storage uses local JSON in dev and Vercel KV in production.
MIT — use it, fork it, ship it.
Halsey Huth — building in public.
Powered by OpenClaw.
If Shipyard OS helps you run your agent team, give it a ⭐