← All terms

Restore Point

A saved snapshot of an AI agent's state at a specific point in time that it can be reverted to.

Restore Point

A restore point is a saved snapshot of a system's state at a specific point in time, captured so that the system can later be reverted back to exactly that state. In the context of an AI agent, a restore point typically captures the agent's memory, configuration, and any persisted data as it existed at the moment the snapshot was taken, forming one entry in a timeline of recoverable states.

How it works

Restore points are usually created either on a schedule, on demand by an operator, or automatically around significant events, such as the completion of a unit of work. Each restore point is stored independently and labeled with when it was taken, so an operator browsing an agent's history can see a sequence of states over time rather than only the current one. Reverting to a restore point, a rollback, replaces the agent's active state with the one captured at that point.

Why it matters

Long-running agents can drift into an undesirable state through a bad task, a misconfiguration, or unexpected input, and without restore points the only way back is to recreate the agent from scratch, losing everything it had accumulated. A timeline of restore points gives an operator fine-grained recovery options: rather than an all-or-nothing choice between the current state and a full reset, they can pick the most recent point before things went wrong and revert to exactly that.

Restore points vs backups

A restore point is similar in spirit to a backup, but it is usually tied more tightly into the normal operation of the system rather than kept separately for disaster recovery. Restore points are typically created frequently and automatically as part of ordinary use, such as after each unit of work, and are meant to be browsed and reverted to routinely, not only pulled out in an emergency.

Restore points in Agenhood

Agenhood takes an automatic snapshot of every agent after each task it completes, building a restore-point timeline without requiring any manual action from the operator. Because rollback in Agenhood is non-destructive, reverting to an earlier restore point does not discard the state that existed before the rollback; that state is preserved as its own point in the timeline, so a rollback can be reversed later if needed.

Get started

Deploy your fleet.

Put a fleet of sandboxed agents to work on your own infrastructure, provisioned in seconds and watched live from one console.

Get started

Admin-provisioned · Self-host in one command · Your data never leaves your VM