Function run_loop 

pub fn run_loop(
    intel: &IntelClient,
    servers: &[McpClient],
    input: &LoopInput,
    self_handler: &mut dyn SelfHandler,
    log: &Logger,
) -> Result<(Outcome, Usage), LoopAbort>

The agentic loop over explicit inputs — one session, one turn. Used by once-mode (run_root) and a per-event subagent run (subagent::control). self_handler supplies agentd’s in-process self-tools (e.g. subagent.spawn); the loop tries it before MCP. A warm continue-session instead drives Session directly across many turns.

Returns the run’s Outcome together with the run’s total token Usage. A one-shot run is exactly one turn, so the run total IS that turn’s usage, and the control layer emits it once per run as a single crate::subagent::protocol::AgentMsg::Usage. Emitting both a cumulative and a per-turn figure would double-count against the tree’s token ceiling.