How Approvals Work in FuseIQ
Understand risk badges, approve/reject flows, auto-approval rules, and audit logs for AI agent governance.
FuseIQ's governance system puts you in control. Every agent action can be assigned a risk level, and you decide which actions require human approval before execution.
Risk Badges
Each pending action in the approval queue displays a risk badge:
Read-only queries, data retrieval, status checks. Green dot indicator. Auto-approved by default.
Data transformations, API calls to known endpoints, content generation. Yellow dot indicator. Requires approval or auto-rule match.
Financial transactions, external API writes, data deletion, agent deployment changes. Red dot indicator. Always requires manual approval.
Production deployments, billing changes, security-sensitive operations. Red dot with intensity animation. Always requires manual approval from a Director.
Approval Queue
The Approvals dashboard displays a queue of pending requests. Each item shows:
- Requesting agent name
- Action type (e.g., "Deploy Workflow", "Execute Payment")
- Risk level badge (Low/Medium/High/Critical)
- Timestamp of request
- Action payload preview
- Approve and Reject buttons
Auto-Approval Rules
Reduce approval fatigue by creating auto-approval rules:
- By Agent — Trusted agents auto-approved for specific actions
- By Action Type — E.g., all "read" actions auto-approved
- By Time Window — Auto-approve during business hours, require approval after hours
- By Cost Threshold — Actions under a set cost auto-approved, above requires review
- By Content Pattern — Regex match on the action payload
Audit Logs
Every approval, rejection, and auto-approval is logged in the Audit Log. Logs include timestamp, requesting agent, action details, approver, and outcome. Audit logs are immutable and retained for 90 days on all paid plans.