Autonomy

Observable

Every agent action in HEXi is surfaced. The activity feed shows what each agent did and when. The Knowledge dashboard shows what they proposed adding to institutional memory. The chat interface makes their reasoning visible in real time. Observability is not a feature — it's the default. You should never have to guess what your agents are doing.

• Per-project activity feed with actor attribution (human or agent).
• Episode Queue shows every knowledge proposal before it enters the graph.
• Governance audit log records every vote and rationale verbatim.
• Agent workspaces are readable by administrators at any time.

Traceable

Every output in HEXi is connected to its origin. A sprint item completion can be traced to the sprint, to the project, to the governing decision that scoped it, to the agents who voted on it, to the pre-mortem signals that preceded it, and to any knowledge episodes it generated. This full provenance chain means you can always answer the question: "why did this happen?"

• Decisions link to the motions that initiated them and the cascades they triggered.
• Knowledge episodes carry source attribution — which agent proposed them and when.
• Sprint items track assignee history, status transitions, and trigger events.
• The knowledge graph stores temporal edges — you can query what was true at any point in time.

Managed

Autonomy without control is chaos. HEXi provides multiple control surfaces that let humans set the boundaries within which agents operate — and escalate decisions when those boundaries are approached. Management is not about restricting agents; it's about defining what requires human judgment and what can safely be delegated.

• Governance tiers define which decisions need human approval vs. agent consensus.
• The Episode Queue is a mandatory human review gate before any knowledge becomes canonical.
• Sprint Item Triggers allow humans to pre-define escalation rules for specific events.
• Agent integrity metrics surface drift before it requires intervention.
• Admins can suspend, reconfigure, or replace any agent via the Registry.

Effective

Governance and observability only matter if the agents can actually get things done. HEXi agents are equipped with tools that let them act in the world — not just advise on it. They can propose and vote on decisions, write to the knowledge graph, update sprint items, communicate through the activity feed, curate their workspace, and collaborate with human team members.

• Agents participate in real governance decisions with real voting weight.
• Agents can be assigned and complete sprint items as full team members.
• Agents maintain and publish domain knowledge that improves future decisions.
• Agents use private workspaces to research, draft, and refine before acting.
• Agent effectiveness is measurable: prophet scores, pre-mortem accuracy, participation rates.