Straight answers for deployment, runtime and control.

AgentOS is an open-source operating layer for building and running teams of digital workers. It organizes workspaces, context, tasks, model access and human supervision around the OpenClaw runtime.

AgentOS supports a trusted local operator setup and an explicitly configured hosted deployment. The published Railway template is a documented hosted path.

Railway provides the published template path for running AgentOS and its pinned OpenClaw Gateway together with a persistent volume. You can also use the local installer for a trusted operator machine.

Railway infrastructure usage is billed by Railway. Model usage is billed separately by the provider you connect. AgentOS is designed for bring-your-own-key provider access.

Yes. Connect a model provider through AgentOS Setup Center and verify a ready model before dispatching work. Keep provider credentials out of public logs and issue reports.

The published template uses a persistent `/data` volume for AgentOS and OpenClaw state across redeploys. The specific data processed by models, accounts and connected services is also subject to those providers’ configurations and policies.

Browser profiles and account targets can be surfaced through the OpenClaw runtime. The current integration is an MVP bridge and typed browser-profile dispatch is not yet exposed by OpenClaw.

Railway can restart a failed service according to its configured restart policy. With the published volume in place, persisted AgentOS and OpenClaw state survives redeploys; inspect deployment logs and runtime health after recovery.

Use `agentos status` and `agentos doctor --deep` from a trusted local environment. Check model readiness, Gateway reachability and authentication before assigning real work.

AgentOS keeps unsupported controls visible. Account access rules requiring approval are intentionally blocked until approval dispatch exists; do not treat them as a completed feature.