Properties vs resources
A device exposes two kinds of things, and knowing which is which is the single most useful distinction on the platform.
- A property is stored state, kept on the platform in the device record. Reading it talks to the platform.
- A resource is a live operation the agent runs while connected. Calling it talks to the device, right now.
If the device is offline you can still read its last properties; you cannot call its resources.
Properties
A property is a named value (any JSON shape: a string, a number, an object) persisted on the device record. It survives restarts and disconnections, and it's visible from the dashboard.
Use properties for state that should outlive a session: configuration, reported firmware version, the site a device sits in, a customer reference, the result of the last maintenance run.
thinr device property list edge-gw-17 # list this device's properties
thinr device property get edge-gw-17 firmware # read one
thinr device property get edge-gw-17 firmware -f value.version # drill into a sub-field
thinr device property set edge-gw-17 site '{"city":"Madrid"}' # write oneset replaces the whole value (it's parsed as JSON, and the property is created when missing). The web console and AI assistants write properties too: over MCP it's type device_property. Persisting structured state this way is the right alternative to running a command and parsing its output every time you need it.
Resources
A resource is a typed endpoint the agent registers while it's connected: it runs something on the device and returns a result. The agent ships built-in resources (like monitoring), and every custom script you add becomes one too.
Use resources for actions and live reads: restart a service, read a sensor, run a diagnostic, query the current disk usage.
thinr device resource list edge-gw-17 # list resources, with their schemas
thinr device resource call edge-gw-17 monitoring # call one
thinr device resource call edge-gw-17 update -s action=check # call with typed inputBecause resources are typed (they can advertise an input and output schema), the web console renders proper form fields for them and AI assistants can call them safely.
Which one?
Ask whether you're storing a fact or triggering behaviour.
| You want to… | Use a |
|---|---|
| Remember which firmware a device reported | property |
| Record the site or customer a device belongs to | property |
| Restart a service or reboot | resource |
| Read a live sensor or current metric | resource |
| Expose "cash level" so you can chart and alarm on it | resource (custom script) |
A useful pattern combines both: a resource computes or fetches a live value, and you persist the important results as properties so they're queryable across the fleet even when devices are offline. Both scale to whole fleets through products.