Spatial infrastructure for XR apps, without running a backend
ReactVision Spatial gives your app persistent anchors, shared multi-user spaces, spatial intelligence, and hosted 3D content as managed services. Your agent wires it up through the MCP, and your users get AR that remembers where everything is.
Anchors, shared state, and hosted content are managed services. Your app passes an API key and a project id.
iPhone, Android, Meta Quest, and Apple Vision Pro share one coordinate frame in real time.
Start on the free tier, or run Spatial on-premises where data, anchors, and assets never leave.
What is ReactVision Spatial?
ReactVision Spatial is managed spatial infrastructure for AR and VR apps: Cloud Anchors, Geospatial Anchors, Co-Location, replicated multi-user state, AI asset generation, asset hosting with over-the-air updates, and environment rendering, plus Spatial Intelligence (VPS, object detection, semantics, depth, and surface understanding). It is built into ViroReact, so there is no backend for you to build or run.
Most AR stacks give you a renderer and leave persistence, multi-user sync, and content delivery to you. ReactVision owns the whole stack, from the ViroReact renderer to the spatial services underneath it, so the pieces are built to fit and your agent has one set of APIs to learn. Comparing hosted anchor providers? See how Spatial stacks up as a Niantic Lightship alternative.
- Reaches your app
- Through ViroReact, on the scene navigator you already use
- Credentials
- An rvApiKey and rvProjectId from your free ReactVision account
- Backend
- Managed by ReactVision, or deployed on your own servers
- Licence
- Commercial service. ViroReact stays MIT and free forever
Everything a spatial app needs after the renderer
Seven managed services, each usable on its own. Use one, or use them all from the same project and the same API key.
Cloud Anchors
Persist content at a physical location across sessions, users, and devices. Place a model in a room today and find it in the same spot tomorrow, on someone else's phone.
A maintenance note pinned to one valve, found by the next technician on a different device.
Geospatial Anchors
Pin content to latitude, longitude, and altitude. Drop a 3D model on a street corner or any real-world coordinate and it is there every time.
A city trail of AR waypoints at real street corners.
Co-Location
Multiple people and devices share one coordinate frame in real time, across iOS, Android, Meta Quest, and Apple Vision Pro. Everyone sees the same content in the same place.
A training team on Quest headsets and iPads gathered around one virtual engine.
Replicated state
Server-authoritative shared state for multi-user apps. Every peer sees changes in the same order, ownership decides who can move what, and state survives the room emptying.
A tabletop game where only the player holding a piece can move it.
Asset generation
Generate 3D models and images with AI, straight into your project. Your agent can do it from the same session it writes your code in.
A prompt becomes a textured prop, placed in the scene a minute later.
Asset hosting and OTA updates
Stream scenes and 3D assets into your app at runtime. Host your team's content and your users' uploads separately, and update content over the air without shipping a new app.
A retail app swaps its seasonal product models with no app store release.
Environment rendering
Render AR scenes inside recorded real-world rooms, with no device in hand. Record a space once, then test the same scene across as many rooms as you like.
Your agent checks a furniture placement in three recorded living rooms before anyone picks up a phone.
Building something multi-user? Read building multi-user experiences with Co-Location, or which anchor type your app needs. Co-Location and replicated state power collaborative MR and VR training, and anchors with a free tier are why startups build persistent, location-based products on Spatial.
Spatial Intelligence: apps that understand the room
Anchors remember where things are. Spatial Intelligence tells your app what is around it: where it is, what it can see, how far away things are, and which surfaces it can use. It is part of ReactVision Spatial, available through the same ViroReact APIs.
Visual Positioning (VPS)
Scan and host a space, relocalise any device inside it, and attach content to its mesh. Content lands where you put it, to the room, not to GPS.
On-device object detection
YOLOE runs through ONNX Runtime on the device itself, with no cloud round-trip. Each detection reports a label, a confidence, and where it is.
Semantic understanding
Know what the camera is looking at: sky, building, tree, road, person, and water. Mask, place, or react to content by what is in the scene.
Depth and world mesh
LiDAR depth and a world mesh where the hardware has them, with a monocular depth fallback, so virtual content is occluded by real objects and collides with real surfaces.
Surface understanding
Horizontal and vertical surfaces with classification, including Meta Quest room models, so content sits on the floor, the table, or the wall it belongs on.
Advanced VPS
The next generation of ReactVision's Visual Positioning System is in development. We will announce it on the Updates blog when it ships.
See how it fits together in how ViroReact lets AR understand space. Mapping buildings with VPS? See spatial data for real estate and proptech.
Built for agents
Spatial features are the part of an XR app agents get wrong most often, because the APIs move fast and the bugs only show up in a real room. The ReactVision MCP Server gives your agent the current APIs, AI asset generation, and environment rendering, so it can build spatial features and prove they work.
Your agent learns the live APIs
Through the ReactVision MCP Server, your agent reads the current ViroReact APIs for anchors, geospatial, co-location, and hosted scenes, and checks which features each device supports before it writes a line.
It generates and hosts the content
The agent generates 3D models and images with AI and adds them to your project, so the assets your app streams at runtime are created in the same session as the code that loads them.
It tests the result in real rooms
Environment rendering lets the agent run your scene inside recorded real-world spaces and compare the renders, so spatial bugs are caught before a person ever holds a device.
Not writing code? The ReactVision Editor puts anchors, hosted assets, and AI generation in a visual editor with its own AI agent.
Deploy it where you need it
Start on our managed cloud, or run ReactVision Spatial and Spatial Intelligence inside your own secure infrastructure. The same app code works against either.
| Option | Managed cloud | On-premises |
|---|---|---|
| Where it runs | ReactVision's managed cloud | Your own secure infrastructure, inside your network |
| Your data | Hosted and managed by ReactVision | Data, anchors, and assets never leave your infrastructure |
| What is included | All of ReactVision Spatial, including Spatial Intelligence | All of ReactVision Spatial, including Spatial Intelligence |
| Pricing | Usage-based, with a free tier | Fixed licence plus maintenance |
| SSO and audit logging | On enterprise plans | On enterprise plans |
| Custom hardware | ViroCore brings ViroReact to your device | ViroCore brings ViroReact to your device |
| Lock-in | None: no device or vendor lock-in | None: no device or vendor lock-in |
Start free, pay as you grow
Sign up for free, no card required. Every plan includes ReactVision Spatial, the Editor, the MCP Server, and ViroReact. You only pay when your asset hosting, API, or AI usage grows beyond the free tier.
Free
- 500 MB asset hosting
- 1,000 Spatial API requests / mo
- 350 AI credits / mo
- MCP server, Editor, and ViroReact
- Powered by ReactVision watermark
- No Environment Testing
Basic
- 1 GB asset hosting
- 10,000 Spatial API requests / mo
- 1,000 AI credits / mo
- No watermark
- Environment Testing
- Everything in Free
Pro
- 5 GB asset hosting
- 100,000 Spatial API requests / mo
- 2,500 AI credits / mo
- No watermark
- Environment Testing
- Everything in Basic
Enterprise
- Custom limits for high-demand projects
- Option of a dedicated ReactVision Spatial instance
- Managed cloud or on-premises
- Custom hardware via ViroCore
- No device or vendor lock-in
- SSO and audit logging
- Direct support from the ReactVision team
- Everything in Pro
Asset hosting covers 3D models, images, videos, and audio, across both your team's content and assets your users upload. Spatial API requests cover Cloud Anchor, Geospatial Anchor, asset storage, and Editor scene calls via ViroReact. Co-Location requires a paid plan. AI credits cover asset generation, the Editor's AI agent, and MCP usage. ViroReact itself stays free under MIT, regardless of plan. Need custom limits, SSO, or on-premises? See Enterprise.
ReactVision Spatial questions, answered
What is ReactVision Spatial?
Do I need to run my own backend to use ReactVision Spatial?
Which devices does ReactVision Spatial work on?
What is the difference between Cloud Anchors and Geospatial Anchors?
Can AI coding agents use ReactVision Spatial?
Can ReactVision Spatial run on-premises?
How much does ReactVision Spatial cost?
Give your app a memory for places
Connect your agent and let it build with anchors, co-location, and hosted content, or create a free account and start building yourself. No backend, no card required.