Save
The agent saves the active document and verifies the expected artifact path.
Recovery controls
Protect the artifact, then make the move.
Adept AI’s proposed checkpoint flow is explicit: save in the application, capture a safe artifact state, verify persistence, and restore only when needed.
Discuss your use caseA recovery point is more than a screenshot.
The recoverable unit is the saved artifact and its required project files. In-memory edits must first be committed by the desktop application. Checkpoints carry identity, creation time, safety status, and persistence outcome so an agent can decide whether recovery is trustworthy.
- Create checkpoints before high-impact actions
- List live and retained recovery points
- Download a selected artifact bundle when available
- Restart without recovery or restore the latest safe point
- Resume stopped work on fresh capacity
Artifact timeline Restore selected
14:02:18Checkpoint #016safe14:08:41Application saveverified14:10:05Checkpoint #018selected14:11:26Unexpected dialogcontainedPersist
The platform records whether the checkpoint reached durable storage; HTTP success alone is not treated as proof.
Recover
The latest safe artifact is restored and the application is restarted, then observed before work continues.
These guardrails reflect public reliability patterns documented for Isle checkpoints and recovery (Python SDK; REST API). Adept AI’s implementation remains pre-release.
Early access
Plan with us before launch.
Tell us which applications, workloads, and agent stack you need. Early access is selected and subject to technical fit and capacity.