Build a behaviour-triggered webinar email sequence by prompting an AI assistant
Describe who gets which email after watching how much, and an AI assistant builds every WebinarStack rule in one call. The prompt to paste, and the limits.
You can build a whole behaviour-triggered email sequence for a WebinarStack webinar in one prompt. Describe the rules in plain language, meaning who gets what after watching how much, and the assistant calls configure_automation_sequence. That tool creates every rule in one all-or-nothing call, across 15 trigger types including watched minutes, watched percent, poll response and CTA click.
This page builds one funnel from start to finish and gives you the prompt to copy. The wiring is a prompt now. The email copy is still yours to write, and the last section says why.
What sequence are we going to build?
Five emails across five triggers, which is enough to cover a webinar from registration to follow-up:
- A confirmation the moment someone registers.
- A reminder an hour before their session.
- A "you left before the offer" email to anyone who leaves having watched less than 60 percent.
- A "you watched the whole thing" email ten minutes after someone finishes the video.
- A replay email an hour after someone is marked as a no-show.
Rules 1, 2 and 5 are plain lifecycle emails. Rules 3 and 4 are the behaviour-driven ones, and they are the reason to do this in WebinarStack rather than in a generic email tool. The watch-time email article covers what to send at each threshold. This page covers how to get it built.
What do you need before you start?
You need an MCP key, an assistant that can connect to it, and a webinar that already exists. The WebinarStack MCP server guide covers creating the key and connecting Claude or another client. The key needs the automations write scope, and nothing else is required for this job. If you want the assistant to read the webinar first, add the read scope.
You do not need the webinar to be finished. Automation rules can be created on a draft, and you can build the sequence before you have written the registration page.
What is the prompt?
Paste this into a connected assistant, replacing the bracketed parts. It is written the way you would brief a person, because that is enough:
Use WebinarStack to add an email sequence to my webinar "[webinar title]".
Create these five email rules in a single call:
1. When someone registers, email them immediately. Subject: "You're in:
[webinar title]". Include the watch link.
2. One hour before their session starts, email a reminder with the watch link.
3. When someone leaves the webinar and has watched less than 60% of the video,
email them 15 minutes later. Subject: "You missed the best part". Point them
to the replay and mention [the offer] is covered in the last third.
4. Ten minutes after someone finishes the video, email them the offer:
[one-line offer description] and the link [offer URL].
5. An hour after someone is marked as a no-show, email them the replay link.
Write the subject and body for each. Use WebinarStack's personalization
tokens for the attendee's name, the webinar title and the watch link.
Before you create anything, show me the five rules as a list.
After you create them, read the sequence back to me.
The last two lines matter. Asking for the plan first gives you a chance to catch a misread rule before it is written, and asking for a read-back means you see what the server stored rather than what the assistant remembers sending.
What does the assistant do with it?
It calls configure_automation_sequence once, with an array of rules. Each rule carries a name, a trigger type, an action type, a delay in minutes, optional conditions, and the action's settings: for an email, a subject and body. A single call accepts between 1 and 20 rules.
The call is all-or-zero. Either every rule is created or none is, so a malformed fifth rule does not leave you with four orphaned rules and a sequence that quietly stops halfway. If one rule fails validation, the whole step rolls back, the tool returns an error, and nothing is written. You fix the one rule and send the set again.
The behaviour-driven rules from the funnel look like this when the assistant sends them:
- Rule 3 uses the
lefttrigger with a delay of 15 minutes and one condition:watched_percentless than 60. - Rule 4 uses the
video_completetrigger with a delay of 10 minutes.
Conditions are a flat list combined with AND. Each is a field, an operator and a value, and the operators are ==, !=, >=, <=, >, < and contains. There is no OR and no grouping. If you want either of two conditions to send an email, you create two rules.
Which triggers can you use?
An assistant building rules over the MCP server can use 15 trigger types. This is the complete list:
registered, joined, left, no_show, watched_minutes, watched_percent, video_complete, cta_click, poll_response, time_elapsed, before_session, after_session, branch_taken, segment_completed and chat_response.
The ones you will use most:
registeredfires when someone registers. It is where the confirmation email goes.before_sessionandafter_sessionfire relative to the session start and end. The delay says how far before or after.watched_minutesandwatched_percentfire once an attendee has watched at least a threshold, while they are still in the session. The threshold is a condition on the rule, for examplewatched_percent >= 50.video_completefires when the attendee reaches the end of the video.leftandno_showcover the people who did not stay or did not arrive.cta_clickfires when an attendee clicks a call to action on the timeline.poll_responsefires when an attendee answers a poll. The interactions tutorial shows how to place one.branch_takenfires when an attendee is routed down a branch of the webinar.
The product itself has 18 triggers. The other three, live_complete, replay_started and replay_complete, separate a live finish from a replay finish. No MCP tool that creates a rule accepts them, so an assistant cannot build one from scratch for you. They are available in the dashboard and over the REST API, and automations_update will accept one on a rule that already exists. If your sequence turns on that distinction, build those rules in the dashboard.
What actions can a rule take?
Four: email, webhook, integration and meta_capi. A rule on any of the 15 triggers can send an email, call a webhook URL, run an action in a connected email tool, or send a Meta Conversions API event. The webhook article covers the webhook version of a watch-time rule.
One constraint to know about: meta_capi rules need the Growth plan and a Meta pixel connected to your account. The assistant looks up the pixel with the meta_pixels_list tool, and the rule refers to the pixel's id from that list.
SMS is not on the list. The dashboard can send SMS from an automation, but SMS is deliberately not available through MCP or the REST API, so an assistant cannot create it. If you ask for a text message, a well-behaved assistant will tell you so.
How do you check and fix what it built?
Ask the assistant to read the sequence back, which it does with automations_list. The list comes back in lifecycle order, not creation order: registered first, then the in-session triggers, then before_session, after_session and no_show last. Rules with the same trigger sort by delay. That means the read-back reads like a funnel, whatever order the assistant created the rules in.
Two more tools help when you iterate:
automations_find_matchingchecks whether the webinar already has a rule with the same trigger, action and conditions. A well-behaved assistant calls it before adding a rule, so asking twice for the same email does not create a second copy that emails your registrants twice. It is guidance the assistant follows, not a lock the server enforces, so read the list after any change.automations_updatechanges one rule, so you can tweak a delay or a subject without recreating the sequence.
There is also an automations_delete tool. It asks the assistant to disable a rule rather than delete it where that would do, and it returns a snapshot of what was removed.
Where does a human still have to look?
The email copy. An assistant will happily write five emails, and they will read like five emails an assistant wrote. They will be correct about the triggers and generic about your audience, your offer and your voice.
Treat the generated copy as a first draft. The product ships an authoring guide for automation emails as an MCP resource, webinarstack.docs.automation_email_authoring, and most clients can attach it to the conversation so the assistant writes against it. That improves the first draft. It does not replace reading it.
The claim this page is willing to make is narrow: the wiring of a behaviour-triggered sequence is now a prompt, and the writing is still yours. Check the thresholds against your own webinar's length and where your offer actually appears, because an assistant cannot watch your video and tell you that 60 percent is the right line.
For the rest of what the platform does, see the features page. For how it compares with an automation layer bolted onto Zoom, see WebinarStack vs AEvent. To place the polls and CTAs these rules react to, see adding timed polls and CTAs with an AI assistant.