Update — October 14, 2026
Protect this app and Agent 0.6.32
My Apps now guides protection for a running Memos 0.31.0 or Uptime Kuma 2.5.5 app with its unchanged reviewed recipe and an owned hostname. The destination must be an online compatible server on the same account in another reported country. The declared country of an external server remains unverified.
Confirm three distinct operations: create the unrouted standby, pair the deployments and run the first restore drill. Opening or resuming the assistant does not authorize a write. It neither purchases a server nor changes traffic. The hosted connector and local MCP 0.50.0 or later expose the impreza_protect_app prompt; local MCP 0.51.0 is the current published package. Use the separate reviewed path for custom image, Git or Compose projects.
A verified drill records a dated copy and its measured restore duration. Copy age still matters: later writes may be lost, and a traffic cutover can interrupt service. A queued or failed drill is not verified protection.
Agent 0.6.32 preserves restore waiting deadlines across restarts, reports a restore job that disappears and blocks sensitive operations when recovery cannot be verified. Health checks, logs and moving traffic away remain available. Finished app-backup work folders are cleaned up after the result is acknowledged; S3 backups and retained restore data are not deleted by that cleanup. Update existing Agents explicitly and check the operation's result.
Follow the recovery preparation guide, Read the backup and restore guide, or consult the technical assistant reference.
Update — October 11, 2026
Jurisdiction Failover now includes eligible custom projects deployed from an image, Git or Compose. Create an unrouted second deployment with standby: true using the same pinned source. Declare all persistent data binds and named volumes; move configuration and secret references through export, review and apply.
Pair the deployments, verify a restored completed backup and complete a verified drill for that exact pair before cutover. Missing data coverage or a missing drill is refused. The current hosted MCP and local MCP 0.49.0 expose this path.
Cold-copy age, review expiry, account permissions, managed-hostname eligibility and human confirmation still apply. No zero-loss, zero-interruption, arbitrary live-database or automatic-promotion guarantee is introduced. Read the current technical reference or the customer guide.
The original October 9 announcement below records the initial catalog release.
Update — October 10, 2026
The reviewed cold-standby catalog now includes Memos 0.31.0 and Uptime Kuma 2.5.5. Use the same app and reviewed version on both sides. The hosted connector supports Kuma, and local clients need impreza-mcp 0.48.1 or later. The public REST reference now documents standby, sync, cutover, failback and drill operations. Existing ownership, permissions, source-fencing and explicit-confirmation requirements still apply.
Agent 0.6.30 also improves the separate, eligible custom-app redeploy workflow: it switches prepared versions without reloading the managed proxy, preserves the previous Git build context if its exchange fails, and coordinates egress-rule updates across processes. These redeploy improvements do not make cold failover interruption-free or loss-free. Existing servers receive them only after an explicit Agent update.
After a Kuma cutover, verify the dashboard, restored monitors and notification channels. The current failover limits, including backup age and managed-hostname eligibility, remain in force. Read the updated customer guide.
The original October 9 announcement below describes the initial release.
Jurisdiction Failover is now available through My Apps, REST and supported MCP connections. The first catalog standby workflow uses Memos 0.31.0 on two owned servers in different reported countries, with an Impreza-managed HTTPS hostname.
Prepare and verify the copy
- Create a healthy Memos 0.31.0 primary with its managed hostname.
- On the second eligible server, select Create a cold standby. Domain and onion inputs are disabled; standby creation does not copy data.
- Pair the primary and standby in cold mode. Back up the primary, restore that exact completed backup to the standby and wait for its healthy redeploy.
- Confirm sync with the exact backup, restore and deploy-command receipts. A ready label is revalidated before preparation; the current workflow refuses a cold copy older than one hour.
In MCP, impreza_deploy_catalog_app supports standby: true for the reviewed build. Omit domain and onion arguments. This option is available in the hosted connector and local impreza-mcp 0.48.0. The reference documents the creation and pairing calls.
Review, confirm and follow the cutover
Preparation returns an expiring review of identities, countries, hostname, backup age and interruption. It changes no traffic. After the responsible person confirms the exact proposal, application requires its unchanged review digest and explicit confirmation.
The source is fenced before target activation to prevent two active writers. Keep reading the saved cutover until verified or an actionable failure. Verification covers the fence, route ownership, target activation and probes. The managed DNS record changes and the target obtains HTTPS; resolver caches can delay what visitors observe.
An unreachable source is not bypassed by an automatic promotion. If activation fails after fencing, inspect the receipt and use the separately confirmed retry path where applicable. A return requires a separate fence-release review, reversed pair, fresh verified copy and ordinary return cutover.
Eligibility and recovery boundaries
- This is a backup-based cold copy: writes after its recovery point remain at risk. Neither a drill nor a verified cutover promises zero loss or zero interruption.
- Customer domains, private onions, arbitrary catalog builds and automatic health-based promotion are outside this initial workflow. A supported public onion can accompany the managed hostname.
- External-server countries are owner-declared, not verified by Impreza. Managed VPS services registered with a public IPv6 Agent address are temporarily excluded from the available server list.
- Use the required account-level permissions. The named hosted workflow requires an account credential without resource confinement; reading, preparing and applying use different scopes.
- Pairing and promotion do not buy servers or change billing. Retirement of the old managed VPS follows its verified termination through the normal service procedure; the retirement action itself does not cancel that service or erase backups.
Drills refresh and verify the standby while traffic stays on the primary. Reported drill durations exclude human review and do not measure the duration of a DNS cutover. Recovery objectives remain goals, not a service guarantee.
Recent supporting updates
MCP 0.48.0 adds the reviewed failover tools, VPS capacity recommendations, pairing local access by code and structured error guidance. Check the current interface-specific tool table rather than treating a package total as every account's available tools.
Agent 0.6.29 observes the service for thirty seconds after an update and restores the previous Agent version if it restarts within that window. It also coordinates egress-rule reapplication inside the daemon. Agent 0.6.28 preserves previous Git source when a new clone fails and tightens readiness checks. Existing servers need an explicit update; Agent rollback does not restore application data.
Eligible managed VPS services ordered with No app can now install the Agent from their service page. Capacity recommendations use available measurements to propose an increase and quote; they do not place a resize order automatically.
Read the setup guide, consult the operation reference and service terms, or open My Apps.
