The control deck for Original Intelligence and Artificial Intelligence.
OIDeck is a self-hosted Project Operating System. Humans set the methodology, rules, and objectives. AI agents do the execution. One place to steer work across projects, models, and devices.
Organization
└── Project
├── Objectives · Methodology · Rules · Knowledge
└── Session
├── Device ─ tmux · shell · (docker · k8s)
├── Runtime ─ Tailscale-reachable
└── Model ─ Claude · GPT · Gemini · localA Project Operating System for humans and AI agents.
Organization → Project → Session. Methodology, rules, knowledge, and artifacts live with the project, not the chat. Sessions inherit downward and stay isolated by default.
Run sessions across Claude, GPT, Gemini, or local models. Across your laptop, desktop, or a remote box on your Tailnet. OIDeck schedules the work; the model executes it.
Your data and secrets stay on your machines. FastAPI + SQLite backend, Next.js dashboard, tmux runtime for v0. Open source under MIT.
Coding agents are everywhere. The control plane is missing.
You decide what the AI is allowed to do, what counts as done, and which actions need approval. OIDeck makes those rules first-class objects, scoped from organization down to a single session.
Persistent sessions in tmux keep agents alive across prompts and reboots. OIDeck routes prompts, captures output, and gates production-side actions behind explicit approval.
Sessions attach to branches. PRs are the unit of change. OIDeck wraps the workflow your team already uses instead of inventing a parallel source of truth.
Sessions can cross-check each other and return review, recommendation, or analysis. They never auto-modify another project. Coordination without coupling.
Open source. Self-hosted. Yours to fork.
- LicenseMIT
- BackendPython · FastAPI · SQLAlchemy · SQLite
- FrontendNext.js · TypeScript · Tailwind
- Runtimetmux (v0) · docker / k8s on the roadmap
- NetworkingTailscale-compatible
- StatusEarly access via GitHub
$ git clone https://github.com/jennyruan/oideck $ cd oideck # backend $ cd backend && uv run uvicorn app:app # dashboard $ cd frontend && pnpm dev # device agent $ cd agent && python agent.py register