App events

Many integration nodes can act as triggers. Instead of performing an action, the node listens for an event in the connected service (a new Slack message, a HubSpot contact, a GitHub push) and starts a run each time one arrives.

​

Select a trigger operation

Registration happens as soon as the trigger is part of your workflow, whether you added the node yourself or the AI builder did. You do not have to open the node for it to start listening. Slack, HubSpot, Discord, and Instagram triggers show a Status field naming what they are listening to, and a registration that failed, or never happened, is repaired on its own within a day. A trigger that still has required fields to fill in is not registered yet, whoever is filling them: registration waits until the configuration can run, and happens on the edit that completes it. Changing the node to a different operation, or to one that is not a trigger, removes the registration again.

​

Registered is not the same as receiving

Some services will not deliver events even with the subscription in place. Slack sends channel messages, mentions, reactions, joins, and file shares only to apps that are members of the channel, so a trigger on a channel NoClick was never invited to is registered and still hears nothing. Registration checks for that and puts the answer in the trigger’s Status field. A trigger that will not be fed turns amber and names the reason, and where NoClick can fix it itself the field offers a button:

• Public Slack channel: the status offers Join #channel. Clicking it puts the app in the channel and re-checks, and the status goes back to naming what the trigger listens for.
• Private Slack channel: only a person can add an app, so the status names the /invite command to run in Slack instead, with the app’s handle in it.
• No channel filter: the app has to be in at least one channel, otherwise Slack has nothing to send.

A trigger in that state is also marked incomplete on the canvas and listed on your Dashboard under Needs you, where the row carries the same fix. It does not hold up a run you start yourself. A check NoClick cannot complete, because the service is unreachable or answers with something unexpected, leaves the trigger as it is. A working trigger is never reported broken on a check that could not reach a verdict.

​

Supported events

Instant push events arrive via webhook subscriptions that NoClick creates for you. Google Drive, Google Calendar, and Jira subscriptions expire periodically and are renewed automatically. Discord is the exception: Discord does not send channel messages over a webhook, so On Channel Message and On Bot Mention are delivered over the connection NoClick holds open to Discord, and they need NoClick’s bot to be in the server. Connecting Discord with the Discord Bot Install credential puts it there: you pick the server during the Discord sign-in. Channel on those two triggers also takes a category, which covers every channel under it and the threads opened in them; categories read as Name (category) in the picker, and the list is ordered the way Discord orders the sidebar. Instagram’s On Comment and On Message need the Login connection, which a Facebook system user or Page token cannot stand in for. On Mention is the exception: mentions of your account in other people’s comments and captions arrive through the Facebook Page linked to your Instagram account, so that trigger uses the Facebook login connection instead. See Instagram. Scheduled poll triggers cover services without a usable push API: you pick a schedule on the node (default: every 5 minutes) and each poll fires the workflow only for items it has not seen before. A poll that finds no new items runs nothing downstream, so connected nodes and AI agents stay idle until there is something new. That applies to a run you start yourself as well, so use a node’s mocked output to test the steps after a trigger when there is no new data.

​

What counts as new

A poll trigger takes its starting point the moment it is registered, so anything that arrives after you finish setting it up fires a run, including a test entry you submit right away. Everything already in the form, sheet, or mailbox at that point counts as history and is left alone. Pointing the trigger at something else later, a different form or spreadsheet, takes a fresh starting point for the new resource rather than firing for its entire backlog. Edits that do not change what the trigger watches leave its history untouched. Running the workflow yourself polls the same way a scheduled run does, and anything it finds is delivered to the nodes downstream, including an AI agent wired to the trigger.

​

App events and AI agents

Wire the trigger node directly into an AI agent and each event is delivered as part of the agent’s turn, as an account of what happened rather than the payload the provider sent: the service and the kind of event, who it came from and where, the content, and the IDs the agent needs to act on it. One node cannot both fire runs and provide agent tools: to build an agent that listens and replies in the same service, use two nodes, one with the trigger operation and one wired in as a tool provider. See Channel agents.