Approvals And Needs Action
Agirunner keeps approvals and operator interventions visible instead of burying them inside model prose.
That is an important product decision. If a system expects human judgment, it should expose those moments clearly and with context rather than forcing an operator to reconstruct them from raw logs.
Two Related Surfaces
Section titled “Two Related Surfaces”Operators typically encounter these moments in two places:
- the approval queue and workflow needs-action views
- the selected workflow workbench, where the context for an intervention is visible
What Shows Up Here
Section titled “What Shows Up Here”- approval gates that need an operator decision
- task outputs that need review
- rework or request-changes paths
- escalations from active work
- operator guidance that should be recorded into workflow history
The workbench implementation is structured around a dossier-like summary of:
- what needs a decision
- why it needs action
- what is blocked now
- what work exists so far
- what action is recommended
- what evidence supports the choice
That is a strong product choice. It reduces the chance that an operator has to reconstruct context from raw logs alone.
Why It Matters
Section titled “Why It Matters”This is how Agirunner avoids an all-or-nothing automation posture. The system can run autonomously when it should, but it still gives humans a clean place to approve, reject, redirect, or request changes without breaking the audit trail.
Agirunner also treats operator steering as a first-class workflow action. Guidance can be recorded into the workflow, echoed in history, and used to redirect next steps without losing context.
How It Connects To The Rest Of The System
Section titled “How It Connects To The Rest Of The System”This surface is the join between workflow automation and human intervention:
- Workflows tells you which runs need attention
- Workflow Detail gives the full context for the decision
- Playbooks And Authoring shape where approvals, rework, and operator judgment are likely to show up
- Live Logs and artifacts help an operator validate what actually happened before acting
Without this page, approvals would be scattered across history feeds and task output. With it, the platform exposes intervention as a deliberate part of workflow control.