Handling Action Required items in the Inbox

An AI Playground run can pause partway through when it needs a person to supply information or approve an action, and an agent step in a workflow can pause when it needs an action approved. Orion Pulsars also need you from time to time, to review the changes they propose or to answer a question. CloudRadial AutomationAI collects all of these in the Action Required inbox, where you respond so the work can continue. This article explains the inbox and how to handle each kind of item. It is available to all roles; approving a Pulsar's proposed changes requires the Admin or Owner role.

In this article:

• When a Run Parks
• The Inbox and the Notification Badge
• Input vs. Approval Items
• Providing Requested Input
• Approving or Denying an Action
• Cancelling a Parked Run
• Pulsar Proposals and Questions

When a Run Parks

While an agent works toward a goal, it may reach a point where it cannot proceed on its own. It either needs an answer from you, or it wants to run an action that requires approval. The run pauses and appears in the inbox until you respond, at which point the agent resumes from where it left off. An AI Playground run can park for either reason, and its conversation shows Waiting for you until you respond. An agent step launched from a workflow or playbook runs unattended, so it parks only for approval: rather than stopping to ask for information, it finishes with an answer that names anything it could not find. See the article on the Agent node.

The Inbox and the Notification Badge

Open the inbox from Action Required (/inbox). When one or more items are waiting, a floating Action Required button appears with a red count badge showing how many items await a response, including Pulsar proposals and questions. The button disappears once the last item is resolved. Selecting it opens the same inbox in a side panel so you can respond without leaving your current page.

Each run item shows the run's goal, the originating workflow if any, when it parked, and a prompt describing what the agent needs. A live indicator shows the connection state, and a Refresh action re-loads the list. When nothing is waiting, the inbox reads "Nothing needs your attention right now."

Input vs. Approval Items

Each run item is tagged by what it needs:

• Input: the agent needs information from you before it can continue. Only AI Playground runs raise Input items.
• Approval: the agent wants to run a specific action and needs your approval first.

Pulsar items carry their own tags, Proposals and Question. See Pulsar Proposals and Questions.

Providing Requested Input

For an Input item, read the prompt, type your response in the answer box, and select Send (or press Cmd/Ctrl+Enter). The response is recorded and the run resumes. Once resolved, the item drops off the inbox and the badge count updates.

Approving or Denying an Action

An Approval item shows the proposed tool and, where available, the action details the agent intends to run. Choose one of:

• Approve: let the agent run the proposed action and continue.
• Approve all & future: approve this action and every later action for the rest of this run, so the agent does not pause for approval again on it. AutomationAI asks you to confirm first. This option appears only for AI Playground runs, not for agent steps launched from a workflow.
• Deny: decline the action. The agent continues without performing it.

When the run belongs to an AI Playground conversation, your decision is also recorded in that conversation, with who made it.

Cancelling a Parked Run

If a parked run is no longer needed, select Cancel run on the item. The run is finalized as cancelled and drops off the inbox.

Cancel run works for both AI Playground runs and agent activities launched from a workflow. Cancelling a workflow-launched item also cancels the owning workflow run that is waiting on it, so the workflow does not stay parked. Because this stops any remaining work in that workflow run, AutomationAI asks you to confirm: select Cancel workflow run to go ahead, or Keep run to leave it waiting.

To see the run in context, follow the item's link:

• Open conversation: for a step of an AI Playground conversation
• View run history: for a workflow-launched run
• View audit: for any other AI Playground run, to see its full detail

Pulsar Proposals and Questions

Orion Pulsars are agents that check in on something for you on a schedule. When a Pulsar wants to change something, or needs to know something, its item appears in the inbox above the run items. See Orion Pulsars: agents that keep watch for you.

Proposals

A Proposals item names the Pulsar and how many proposals it has for you to review, when they were made, when they expire, and a short summary. Proposals that are not decided before they expire are not run.

• Review: open all of the Pulsar's proposals in a side panel, where you can approve, deny or edit each one, or use Approve all remaining or Deny all remaining. When every proposal is decided, the approved ones run.
• When there is only one proposal, the item shows its title, the tool and the values it will send, and you can select Approve or Deny right in the inbox.

When you deny a proposal, you can add an optional reason. The Pulsar remembers it and takes it into account in later check-ins. Only Admins and Owners can approve changes; any member can deny with a reason.

Questions

A Question item shows what the Pulsar is asking. Type your reply in Your answer and select Answer. The Pulsar saves your answer and uses it from its next check-in.

If you are still having trouble, we're here to help! Submit a ticket for assistance, and don't forget to check our status page to ensure there are no outages in your area.