Sequel's November 2 product post and the matching Business Wire release, carried the same day by MarTech Series, put Sequel Agents into general availability for every Sequel customer. The surface is three named starters: Build for program planning and pre-event setup, Scale for turning a finished session into clips and copy, and Convert for personal follow-up, seller briefs, CRM leak checks and monthly program reports. Because Sequel runs webinars on your own site, agents can join website behavior with what the same person watched, asked and clicked in-session, then draft work against that combined profile.
The operator problem is narrower. Agents can read and draft on their own. They cannot send email, invite people, change events or push leads to your CRM until someone on your team with the right Sequel role approves. Sequel's pricing page and the GA post both say agent usage stays free until March 1, 2027, then moves to pricing based on completed AI actions. Between now and that meter, the job is naming who may approve each outbound class, which agent jobs stay draft-only on purpose, how MCP write scope is limited, and how you will catch CRM sync failures before a bad batch leaves the queue.
This is a docs-based review. I have not connected a Sequel tenant, run Build, Scale or Convert, approved a batch, or wired Sequel's MCP into Claude. Everything below comes from Kathleen Booth's November 2 launch post, the agents product page, the pricing page, Sequel's Release Hub notes, the Business Wire text as republished by MarTech Series, and Charles Summers' November 4 Hacking Demand hands-on guide. Customer pipeline and asset numbers are vendor or customer claims, not my measurement.
What it is
Sequel Agents are AI agents for webinar and video marketing that run on the first-party audience context Sequel already holds when events live on your website. The agents page frames three lanes:
Build plans and prepares. It reviews attendance and conversion history, recommends series and timing, clones events from a series template when a task is marked ready in Asana or Airtable, sets up registration, CRM campaign and landing page, and runs Event Check 72 and 24 hours before go-live. Problems land in Slack with a suggested fix. Two weeks out, Build can draft 15 to 25 account-level invite packs for reps to send from their own tools.
Scale turns a finished session into a content set. Every morning it picks up events that ended in the last 24 hours, cleans the transcript against your glossary, cuts clips, and drafts a blog, LinkedIn posts for speakers and a newsletter blurb. Recaps can publish to the event landing page as crawlable text on your domain.
Convert works the audience afterward. It drafts follow-up that opens on the question someone asked or the poll they answered. No-shows get takeaways and a replay link. Engaged attendees' signals go toward the CRM once approved, and account owners get Slack briefs that flag open deals tied to attendees. A CRM sync monitor compares registrants with what reached Salesforce, HubSpot, Marketo or other connected systems, flags failures, and reruns after you approve.
Custom agents sit beside the three starters. Describe a repeating job in plain language inside Sequel, or bring events, transcripts and engagement into Claude, Cursor or another MCP client through Sequel's MCP server. Brand Kit, glossary and company memory shape tone. Role-based access mirrors the rest of Sequel.
What changed
Webinar platforms have long ended at broadcast, recording and export. Sequel's bet is that the platform that already joins site behavior to session engagement should also draft the prep, the content set and the follow-up. November 2 is general availability of that agent layer for every customer, not a separate SKU announce. Growth, Business and Enterprise plans all list Build, Scale and Convert as included, with agent usage free until March 1, 2027.
Release Hub notes for the week of November 6 spell out the control surface more tightly than the marketing post. Convert drafts personal follow-up using a standing brief for voice. Repurpose writes clips, cleaned transcript and social or blog drafts into Media Hub. Event Health Check reviews upcoming events for setup gaps. CRM Sync Monitor watches registrations and attendance that never reach connected CRM or MAP systems. Approvals: anything that sends email or changes an event comes for review first; you can edit a batch, exclude recipients and approve; pending approvals stay open for 72 hours. MCP can create, clone and edit events, repurpose recordings, build segments and run agents, but destructive actions ask for confirmation, and email sending stays inside Sequel.
Hacking Demand's November 4 guide reads the same approval fence and suggests Build readiness first, Convert drafts on one event next, Scale once you trust the voice. A bad readiness Slack is embarrassing. A bad approved send is a brand incident.
What works
The approval default is the load-bearing design choice. Draft-without-send lets you turn agents on and inspect output without lighting the audience. Sequel states the four blocked actions clearly across the blog, the agents FAQ and the Business Wire copy: no email send, no invite, no event change, no CRM lead push until a person approves. Release Hub adds the 72-hour pending window and batch edit or exclude. That is a real ops surface, not a slogan.
First-party context is the other load-bearing choice. Agents that only see registration rows invent follow-up. Agents that see pricing-page revisits plus the question someone asked in-session have a better draft to reject or approve. If your webinars still redirect off-site, the agents inherit a thinner picture.
Build's Event Check at 72 and 24 hours is the safest early win. A Slack-only problem list with a suggested fix is useful even if you never approve a Convert batch. CRM Sync Monitor attacks a boring failure mode: registrants who never land in Salesforce or HubSpot. Flagging the miss and waiting for approval before a rerun is the right shape. MCP keeps email send inside Sequel even when connected, which shrinks one class of prompt-injection blast. Treat MCP write scope as a named decision. Free until March 1, 2027 removes the SKU negotiation for a pilot.
What breaks or is undocumented
Customer pipeline numbers are claims, not your proof. ElevenLabs' story in Sequel's launch materials: a $300K webinar pipeline goal, an agentic stack on Sequel with Claude, n8n and Clay, $14M attributed pipeline in the first half. Unit21: Sequel MCP into Claude, about 19 assets from a 60-minute webinar, roughly 3 hours of review versus an estimated 29 hours manual. Sequel's own summit story: BDRs used MCP for nearly 1,000 registrant follow-ups, reviewed a sample of about 20, launched through Apollo, more than $100K net new pipeline in 24 hours. Treat all of that as customer or vendor narrative, not an audit of your funnel.
March pricing is a date without a rate card. Free until March 1, 2027, then pricing based on completed AI actions. No public action definition, no unit price, no plan multipliers, no overage story. Ask Sequel in writing what counts as an action (draft vs approve vs publish vs CRM push), whether rejected drafts still meter, and whether MCP runs bill the same as in-product agents.
Approval coverage needs a checklist, not a press quote. Marketing copy lists send, invite, event change and CRM push. Release Hub stresses email sends and event changes, plus the 72-hour pending window. Confirm in your tenant whether publishing a recap to the event page, writing clips to Media Hub, or posting Slack briefs require the same approval object. Draft-to-Slack and draft-to-site are different blast radii. Write the matrix down.
MCP write scope and format debt. Release Hub says MCP can create, clone and edit events, repurpose recordings, build segments and run agents, with confirmation on destructive actions and email send kept inside Sequel. Map which MCP tools are enabled for which roles before someone "just connects Claude." Sequel supports live, simulive and on-demand; the agents FAQ says not every agent action works the same way in each format. Agents work best with Slack connected, Brand Kit set, company memory saved and tracking pixel installed. Brand corrections are remembered for next runs, so own Brand Kit edits the way you own a style guide change. Hacking Demand corroborates the public fence and the "confirm cost on your plan" caution; do not treat it as Sequel documentation.
Who it's for
Demand gen and webinar program owners already on Sequel who want prep, repurposing and follow-up drafted against real session context, with a named approver on every outbound. Marketing ops leads who will own the approval queue, the 72-hour pending window and the CRM sync failure list. Sales enablement partners who want account briefs tied to what a person asked. Teams already living in Claude or Cursor who want Sequel MCP without letting the assistant send mail directly.
Weaker fit if you are not a Sequel customer, if compliance cannot tolerate any AI-drafted audience email even with human approve, if your events still run on a third-party domain, or if you need a public action price before you will turn anything on. Also weak as a substitute for a general marketing agent suite: the center of gravity is webinar and video programs on Sequel.
Pricing and limits
Sequel Agents are included on Growth, Business and Enterprise. Agent usage is free until March 1, 2027, then priced on completed AI actions. No standalone agent SKU or public action rate card as of this review. Plan floors still follow attendee volume (Growth lists $833/month billed annually for up to 500 attendees per session on the public page; Business and Enterprise are sales-priced). Limits from primary sources: human approval required before email send, invite, event change or CRM lead push; pending approvals open 72 hours; role-based access; email sending stays inside Sequel even via MCP; not every agent action is identical across live, simulive and on-demand. Confirm action metering, rejected-draft billing and MCP parity in writing before you scale Convert across every registrant.
What to do this week
Name the approvers. Write a one-page Sequel agent policy: which roles may approve Convert email batches, event changes from Build, CRM reruns, and whether Slack briefs need a second set of eyes. Use roles, not a rotating personal handle. Pin it next to your webinar runbook.
Turn on Build's Event Check for the next two scheduled events before you approve any audience send. Fix the broken embeds and CRM sync issues it finds. That proves value without touching the audience.
Run Convert as draft-only on one finished event. Compare twenty personal drafts against what your team would have sent. Record edit rate and how often the draft invents a question nobody asked. Only then approve a small segment, never the full registrant list on the first pass.
Open the CRM Sync Monitor after that same event. Assign an owner for the failure queue. Approve one rerun only after the root cause is written down. If you use MCP, inventory enabled tools and roles. Confirm email cannot send from the assistant. Keep MCP write off for guests and contractors.
Ask Sequel five written questions before March: what counts as a billable AI action; whether drafts you reject still meter; MCP versus in-product parity on metering; whether recap publish and Media Hub writes sit behind the same approval object as email; action price bands for your attendee tier. Calendar a February review so March 1 is not a surprise invoice.
Stop treating ElevenLabs' $14M or Unit21's 19 assets as your business case. Those are their claims under their stack. Your case is quieter: who may approve each Sequel send, which agent jobs stay draft-only, how MCP write is scoped, how CRM leaks get caught, and how you will watch the action meter before March 2027 prices hit. Approve on purpose. The free window is for locking that fence, not for unsupervised scale.
