Skip to main content
Cloud Adapter has two deployment modes. Full mode can serve development or production environments. Local development mode runs against your own cloud resources; it is separate from the in-memory emulator.
The container launcher for local development mode is pending as of October 9, 2026. The current local quickstarts use the existing foreground tensor9 adapt svc run command. They do not demonstrate the new container workflow. Tensor9 owns the launcher delivery and will update the quickstarts when that release is available.
Both modes use the same service implementations. A deployment mode does not expand a mapping’s supported API operations. The local emulator uses in-memory backends instead of your cloud services. Acquisition, setup and license renewal require Tensor9 CLI authentication with T9_API_KEY. Local development mode retains the license requirement.

​
Identity and networking

The application must reach the adapter, and the adapter must reach its target services. Give each connection its own network and authentication design. Private placement alone does not authenticate a caller. The planned local launcher exchanges cloud credentials for short-lived credentials and hands them to the container as files. It does not mount your cloud credential home directories. Keep provisioning credentials separate from serving credentials and use the least privilege needed for each operation.

​
State and availability

Inventory application data and adapter state separately. Give each an owner, backup policy and recovery procedure. Stopping a local container does not delete or back up your target-cloud data. For production in full mode, verify the selected service mapping, network controls, backups and failure recovery. A successful local quickstart does not establish production availability or scaling.