Four surfaces, one token, one set of permissions.
| Surface | Use it for |
|---|---|
| API | Anything, from anything that can make an HTTP request |
| CLI | Terminal work and CI pipelines |
| MCP server | Letting an AI assistant work with your workspace |
| Edge apps | Content that no ready-made app covers |
Which one
The API if you are building something that runs unattended: provisioning a fleet from an inventory system, syncing content from a CMS, pulling playback data into a warehouse.
The CLI if a person or a pipeline is driving. It is the fastest way to do something once, and it drops into GitHub Actions without any wrapper code.
The MCP server if the driver is an assistant rather than a script. It exposes the same operations as tools, so “which screens in the Frankfurt plant are offline” is answerable without anyone writing a query.
An edge app when the problem is what goes on the screen rather than how it gets there. Everything above moves content around; an edge app is content.
One token, four surfaces
All four authenticate with the same API token and are scoped by it. A token carries exactly your permissions in one workspace and cannot exceed them, which is the property that matters when you hand one to a pipeline or an assistant.
The dashboard runs on the same API, so there is no capability that exists only by clicking.