The infrastructure layer for physical AI.
PhyStack is the infrastructure layer for physical AI, the system of record for the real world. One platform to deploy intelligence to any device, from a sensor to a robot, connect every system in a space, and give humans and agents a way to act on it.
The few who deploy physical AI today stitch together point solutions: one vendor for the edge OS, another for fleet, a third for analytics, a fourth for device control. Each speaks a different protocol. Each has its own dashboard.
We fix that with:
- Edge-native OS (PhyOS) with OTA + GPU
- Digital twins + MCP for real-world context
- One CLI, API, and Console to manage it all
- Built-in signals + remote ops
- Secure, compliant, multi-region data
PhyStack is everything you need to build for the real world. As easily as you build for the web.
Runs on Intel, ARM, or custom hardware. A secure, GPU-accelerated container OS with OTA updates and built-in failover — perfect for edge AI, high-performance logic, and hardware control.
Your cloud-native backend. It manages digital twins, streams real-time data, powers ML pipelines, and handles everything from identity to products, checkout, and media.
Signals track every event, drive analytics, and fine-tune your models. Spaces map your environment from region to room, so you can target content, apply permissions, and generate rich reporting.
Doors, cameras, printers, sensors, etc — all become controllable through simple APIs and digital twin models. Your apps interact with them just like software objects.
It's all tied together by a complete toolkit for developers: use the CLI, REST API, drop-in SDKs, the intuitive Console UI, or connect agentic workflows via the MCP server.
Build your first edge app in less than 10 minutes:
# Scaffold your edge app
phy app create
# Deploy to a local VM for testing (or connect a device)
sudo phy vm create
sudo phy vm start <vm-name>
# Push and register your app
phy app deploy <device-name>
# Connect to your device or VM shell
phy dev shell <device-name>From here, you can iterate, publish new versions, and deploy at scale — all from the CLI or Console.
Want the full walkthrough? Check out the complete Edge App tutorial.
- Deploy AI models to edge devices. Go from prototype to live edge inference in minutes.
- Run inference and stream analytics. Manage your entire model lifecycle right from PhyStack.
- CLI and SDK support. Automate deployments, scaffold apps, upload models, and bind to devices with Node or Python.
- Roll out to thousands of devices. PhyStack handles orchestration, health checks, rollbacks, and fleet scaling.
- Build adaptive edge apps. Use MCP to make your models environment-aware and responsive to real-world signals.
- Install and manage via Console or CLI. Keep full control over where and how your apps run.
- End-to-end encrypted. Data is secure in transit and at rest (WSS, HTTPS).
- Anonymized user IDs. Easy consent management and withdrawals.
- Global data residency. Choose where your data lives: EU, US, AU, IN, UAE.
- Signed containers with RBAC. Secure deployments and role-based access.
- ISO 27001 certified. Independently audited security and privacy management.
- Firewall and threat detection. Azure WAF, IDPS, and Sentinel SIEM.
- Outbound-only networking. One port to whitelist, secure by default.
It only takes a few steps to get up and running with PhyStack:
- Set up your tenant — register your tenant and create your user account.
- Prepare your dev environment — install the CLI and connect to your tenant.
- Provision your first device — install PhyOS on your hardware and connect your device.
Then head to our tutorials to deploy your first edge app, screen app, or connect them into a complete real-world experience.
Looking for production retail software running on this stack? See PhyGrid
- queue management
- self checkout
- product catalog
- customer engagement for retailers and travel retail.
- and more...