Remote Targets

Premium: remote targets are available with a valid license for the selected configuration.

A remote target names an execution environment for a delegator. Operator keeps the queue and orchestration; the target runs agent processes, which report workflow progress through opr8r callbacks.

Target	Status	Connection
SSH Hosts	Alpha	An SSH alias and remote working directory
Coder	Alpha	Per-ticket Coder workspaces

Local multi-agent execution and local containers remain included. Remote model inference and remote access to the dashboard do not require a remote execution entitlement.

Limitations in this release

Operator enforces these when a delegator resolves to an SSH or Coder target, rather than failing later in the launch:

Constraint	Why
Git worktrees are disabled	The worktree is created where Operator runs, not where the agent does
Relay injection is disabled	The relay hub is a local socket and is not reachable from the target
The Zellij session wrapper is rejected	The launch fails rather than silently running somewhere unexpected

A remote launch therefore works in the checkout at the target’s working directory. Use one target per concurrent piece of work, or separate working directories, rather than relying on worktree isolation.

Targets remain visible when a license expires. Existing work can report completion; a valid license is required to start further remote work.