Trust, Governance & Security · 09.04
Autonomy Modes
Ask, Plan, and Auto, plus deployment + per-workspace autonomy toggles that decide how much a run can do unsupervised.
Layered, not a single global switch
Autonomy isn't one on/off toggle — it's layered. Per-conversation, Ask/Plan/Auto decides how much friction a single chat turn faces. Independently, a per-workspace toggle decides whether background systems like the Command Heartbeat are allowed to act unbidden at all, versus only ever surfacing what needs attention. And self-heal's own guarded vs. bypass mode decides whether structural repair proposes for approval or acts directly. Three genuinely separate dials, each defaulting to the safer setting, rather than one flag that has to somehow mean all three things at once.
Continue
Action attribution with node-level token accounting from a single sink — no anonymous spend.
Every run streams reasoning, a live activity tail back-fills a surface opened mid-run, and a realtime token pill shows spend as it accrues.