MC-1 DOCS
API ReferenceCookbook
Docs/Architecture guides/Forward Autonomy Levels
PublishedGuidanceLAST REVIEWED · 30 AUG 2026

Forward Autonomy Levels

Choose between forward autonomy levels using workload evidence and explicit constraints.

Decision

Use this guide when choosing forward autonomy levels. Neither option is universally superior; choose per workload and keep an escape path.

Choose Lower autonomy

Lower autonomy favors read-only discovery, proposals, and exact human approval.

Choose Higher autonomy

Higher autonomy is appropriate only after identity, policy, sandbox, outcome, and rollback evidence mature.

Material tradeoffs

  • Increase one bounded capability at a time.
  • A higher level is not a goal; use the lowest level that meets the operating outcome.

Validate the choice

  • Test both paths on the same representative set.
  • Compare quality, latency, cost, security, and operability.
  • Document the decision, evidence window, review date, and rollback.
End of guide
Was this page helpful?