Operate Mode
Operate mode is the operator-facing half of the Developer Platform. It serves treasury leads, risk leads, security leads, and platform owners who need to control and audit autonomous agent execution.
Operate Navigation
| Page | Purpose | Status |
|---|---|---|
| Operations Overview | Spend posture, pending approvals, incidents, session health | Public Beta |
| Approval Inbox | Queue actions for review, approve or reject with reasoning | Public Beta |
| Policies | Configure spending limits, chain restrictions, approval thresholds, counterparty rules | Public Beta |
| Incidents | Severity-based incident log with affected agent, session, and timeline | Public Beta |
| Traces | Step-by-step execution log with reasoning, verdicts, signatures | Public Beta |
| Evidence Export | JSON and CSV audit bundles for compliance | Public Beta |
Operator Workflow
- Join or create workspace with operator role
- Review spend posture — active sessions, recent spend, policy defaults
- Configure policies — set limits, chains, approval thresholds, counterparty rules
- Monitor approvals — approve, reject, or let expire with reasoning
- Investigate incidents — review severity, affected entities, related traces
- Inspect traces — step-by-step execution with reasoning hash and signatures
- Export evidence — JSON or CSV bundles for audit
Governance Surfaces
Approval Inbox
Pending actions that exceed policy thresholds appear in the approval queue. Each entry shows:
- Action type and summary
- Amount, tokens, and chain
- Risk score
- Policy that triggered the approval
- Agent and session responsible
Policy Packs
Reusable policy configurations that apply across agents and sessions:
- Spending limits — Per-transaction and daily caps
- Chain restrictions — Allowlisted chains only
- Approval thresholds — Actions above N USD require approval
- Counterparty rules — Approved recipients and protocols
Incident Log
Security and policy violations organized by severity (Critical, High, Medium, Low):
- Title and description
- Affected entity and session
- Related trace hash
- Status (Open, Investigating, Resolved, Dismissed)
Trace Explorer
Per-execution audit trail:
- Agent reasoning (LLM output or intent description)
- Proposed action
- Policy verdicts for each rule evaluated
- Signatures
- Transaction hash (if executed)
- Duration and risk score
Evidence Export
Downloadable evidence bundles for audit and compliance:
- JSON format with full trace, incident, and approval data
- CSV format for spreadsheet analysis
- Filtered by date range, agent, or session
Success Metric
An operator can answer who authorized an action, what policy applied, and which session executed it in under 2 minutes.