App-managed servers

Choose automatically, by subscription, from all manual nodes, or one fixed node. Custom manual groups organize the server list; they do not limit automatic connection candidates. Smart/Custom can use 1–3 entry servers; a fixed node or All via VPN uses one. Eligibility uses ordinary node latency, and background refresh never hot-replaces a running connection.

Choose Smart Routing, All via VPN, or Custom Routing independently. Smart/ordinary Custom expose DNS, sniffing and proxy selection. All via VPN uses only proxy DNS; platform exclusions still determine what enters the tunnel.

Provider and personal configurations

A provider-managed subscription selects one complete Remnawave configuration, without expert mode. Its routing summary replaces the App traffic selector; Smart/Custom preferences remain saved. Provider configurations cannot be combined into App entry/exit pools.

Expert mode selects personal Raw JSON. Opening its browser without selecting a Raw does not immediately change the saved connection target. Server and Raw choices made elsewhere, including the tray, synchronize with Connect.

Runtime and reconnecting

Connect shows current-connection speeds and traffic, not lifetime totals. Visible pages keep sampling even when a desktop window loses focus; hidden windows and other pages pause it. Least-latency multi-chain connections also show probe status, latency and time; probes are not service-unlock checks.

Changes affecting the active configuration require confirmation and reconnection. OneXray stops the old VPN, prepares local assets, then starts the new runtime. Startup does not download missing routing data. Failure reports the actual cause and does not silently restore the old connection. Longer explanations are available from page/group help buttons.

Automatic selection first looks for enough healthy ordinary nodes within the chosen scope. When those candidates are missing or insufficient, it selects the first readable active provider configuration by ID in that same scope. Complete provider configurations are not latency-ranked, combined into entry chains or used as final exits. Cancellation and read, test-service, compilation, permission or startup errors are reported rather than retried through another provider. The saved automatic/source choice remains unchanged.