Agent workspace backends
Compare and safely activate the runtime backend used for new Lifecycle Agent workspaces.
A workspace backend is the runtime where Lifecycle Agent checks out code, runs commands, and serves previews. Administrators select one active backend for new workspaces.
Your Lifecycle release and administrator policy control backend administration. Runtime backends can be unavailable to some users.
Compare backends
| Backend | Necessary connection | Capabilities |
|---|---|---|
| Kubernetes | Lifecycle’s cluster integration | Supports chat, Develop, environment and sandbox sessions, editor, previews, resume, and prewarm |
| OpenSandbox | Workspace image. API key when the service makes it necessary. | Chat and sandbox sessions, previews, and resume. Editor when the image includes code-server. |
| E2B | API key and template | Chat and sandbox sessions, previews, and resume. Editor when the template includes code-server. |
| Modal | Token ID, token secret, and image | Chat and sandbox sessions, previews, and checkpoint-based resume. No editor. |
| Daytona | API key and snapshot | Chat and sandbox sessions, previews, and resume. Editor when the snapshot includes code-server. |
| Substrate | Not configurable | Coming soon |
The remote backends do not support Develop workspaces, Environment sessions, or workspace prewarm. Capabilities can change. The running Settings page shows the current capability data.
Backend configuration and tests
- Open the backend in Runtime backends.
- Enter its necessary image, template, snapshot, endpoint, and credential fields.
- Save the connection.
- Select Test connection to validate credentials and basic configuration.
- Select Launch test sandbox for an end-to-end check.
- Examine each check stage and gateway authentication.
- When the backend supports an editor, examine editor access.
- After the test sandbox succeeds, select Use for new workspaces.
The test sandbox creates and deletes a provider resource. The provider can briefly bill for this resource.
Activating a third-party backend sends repository content, Agent instructions, commands, and runtime data to that provider. Examine its data handling, residency, access controls, network boundary, billing, and deletion guarantees before activation.
Handle credentials
Credential fields do not show stored values. To rotate a credential, replace it. Then, run the connection test and sandbox test again.
Provider environment variables can supply some credentials, but the Settings page does not show their values. Follow your deployment’s secret-management procedure rather than placing credentials in a repository.
Understand activation
After activation, Use for new workspaces affects only new workspaces. Existing workspaces keep their original backend and provider billing.
Before switching:
- read the capability-loss warning in Settings
- complete important work or keep it on the current backend
- use a representative repository on the candidate backend
- make sure that its editor and preview behavior is correct
- record the rollback selection
After switching, start a new test session. Make sure that checkout, command execution, preview access, suspend/resume, and cleanup operate correctly.
See Agent Sessions for the user workflow and Agent administration for the related policy.