Hosted by us
Open the studio. Nothing to configure and nothing to operate. This is the fastest way to see whether the analysis pass tells you anything you did not already know.
Deploy
Running MBE3Dstudio
There are three ways to run MBE3Dstudio (hosted by us, served from your own network, or under a source agreement), and the choice is mostly about who operates it and what it is allowed to reach. The agent is optional in all three.
What you are running
A project is a file, not a record in a service. You open and save *.mbe.json
through the ordinary file picker, and a working copy is kept on the machine you are
using so an accidental reload does not cost you the session.
- No account, no licence key, no telemetry.
- Nothing you model is collected or stored by us.
- The agent is the exception, see the agent below.
The corollary is worth being explicit about: because a project is a file and not a
record in a service, there is no server-side backup. Save your work, and keep it in
version control like any other engineering artefact.
Three ways to run it
Self-hosted on your own network
The application is a folder of static files, so hosting it means putting that folder behind whatever web server you already run: an internal nginx, an S3 bucket, a share. No container and no runtime. This suits environments that may reach an internal host but not the public internet; the agent stays unavailable unless you also run the proxy.
Source access
For teams that need to extend the schema, add a constraint kind, or build MBE3Dstudio into an existing toolchain. MBE3Dstudio is not open source, and source access is arranged directly under agreement. Get in touch with what you are building and what you would need to change.
What it needs
| Browser | Any current Chrome, Edge, Firefox or Safari with WebGL enabled. |
|---|---|
| Graphics | Hardware acceleration is preferable but not required. It runs on a software renderer, more slowly. |
| Network | To load the application, and to reach the agent proxy if you use the agent. |
| Server | Anything that can serve static files over HTTP. The agent additionally needs the Pages Function and an Anthropic API key. |
| Storage | Local storage for the autosave; the filesystem for saved projects. |
Where your data goes
For everything except the agent: nowhere. We do not collect, store or transmit the models you build. A model of a system you are not allowed to discuss over email is not being sent over email. The only copies are the file you saved and the working copy on that machine.
The agent is different, and you should assume it is. Asking it a question sends a briefing about your model to Anthropic's API, by way of a proxy on this site that holds the API key so it is never exposed in the page. A specification you attach travels the same way, section by section, as the agent reads it. If that is not acceptable for your programme, do not use the agent; nothing else in the tool depends on it, and a self-hosted copy without the proxy has no way to reach it.
Clearing site data clears the working copy. Use Save for anything you intend to keep.