Six Weekly Klaviyo MCP Workflows
Generated by Klaviyo AI
Klaviyo’s MCP server can streamline weekly reporting and reveal actionable patterns across campaigns, flows, deliverability, and list growth. The workflows also show how MCP supports analysis while Composer handles campaign implementation.
- Automated reporting: MCP can build weekly reports comparing email, text messaging, flow, and list performance across time periods.
- Retention trends: Analysis by send time, weekday, send position, segment, and subject-line style can identify practical campaign changes.
- Flow audits: Pulling complete flow structures helps uncover channel gaps, mistimed delays, missing logic splits, and stale builds.
Klaviyo's MCP server is one of the most exciting, least-written-about developments in the ecosystem right now. As co-founder of Retention Harbor, an agency specializing in Klaviyo email and SMS for DTC ecommerce, I work with close to 20 DTC brands across apparel, home goods, and accessories, and the MCP server has become part of my weekly rhythm on nearly every one of those accounts. My LLM of choice is Claude, but each of the use cases I describe below would work with other AI tools, as long as you can connect it to Klaviyo via the MCP.
Here are the use cases I lean on most, plus one from a brand-side marketer putting it to work in a different way.
1. The Monday digest: run a weekly flow health check in minutes
Every account I manage gets a weekly check-in, and that used to eat up a few hours every week. Pull the numbers, line them up against last week across email and text messaging, write up what moved, and only then start thinking about what to do with it. Now, MCP builds all of that as a PDF or a slideshow deck, comparing the current state of the account against whatever date I want to measure against: week over week, month over month, year over year.
A few of the metrics I have the integration track every time:
- Weekly email and text message campaign report (top and bottom performers, and why)
- Weekly email and text message flow report (top and bottom performers, and why)
- Weekly list growth report
- Channel report: email vs. text messaging
The prompt I use for this: "Compare this week to last week across email, SMS, and flows: revenue, send volume, open and click rates, conversion rate. Tell me what moved and flag anything that dropped."
2. Spot the retention trends you should actually act on
Every account is sitting on a mountain of rich data. But most of it never gets used, because pulling it apart and graphing the trends manually takes hours. So the patterns that should be steering the next campaign just sit there. MCP does that analysis in minutes instead of an afternoon.
- Send time: I recompute the click rate and revenue per recipient every hour. Audiences routinely engage at a completely different time than when the brand defaults to sending, and evening or late-night windows out-earn the mid-morning slot on the calendar more often than you'd think.
- Day of week: I rank weekdays by revenue per recipient and by click rate separately, because they often disagree. One day carries the week on revenue while another pulls big clicks and weak conversion.
- Send position: On days a brand sends twice, I check whether the second email adds real revenue or just cannibalizes the first. If it's pulling its weight, the double send is justified. If it isn't, that's a send costing money to produce without paying it back.
- Subject-line patterns: I group subject lines by what they're doing (e.g., discount, urgency, curiosity, personalization), and then I look at how each type performs. That way I can lean into the angles that land for that specific list and stop relitigating the ones that don't.
One note here: I always pair subject-line analysis with clicks and revenue, not opens alone, since Apple Mail's privacy changes have made raw open rate a lot noisier than it used to be.
The prompt I use for subject line analysis: "Break my last 90 days down by send hour, day of week, segment, and subject-line style. Show me click rate and revenue per recipient for each, and tell me what I should change."
3. Audit your flow architecture for hidden revenue leaks
A revenue report tells you a flow makes money. It can't tell you the flow is built wrong, and a flow built wrong under-earns subtly, every day, with nothing in the numbers pointing at what's missing.
MCP pulls the full structure of every flow at once, so the audit takes minutes and I can actually go fix what it finds. A few of the leaks I look for most:
- Channel gaps: I check the channel mix of each flow and find the ones running thin. Text messaging often earns considerably more per recipient than email in high-intent moments like cart and check-out abandonment, so a flow catching the hottest traffic on email alone may be underperforming, depending on the list and price point.
- Mistimed delays: Laying every delay out flat makes the bad ones jump out. A post-purchase flow that opens with a one-day delay is sleeping through its best window. The buyer is right there, still on the site, still excited, and the flow is making them wait. Pacing should track intent, too: the hottest triggers deserve the tightest sequences. Bad timing doesn't break a flow, it just lets a working one leave its best moment on the table.
- Missing logic splits: Any flow that can reach an existing customer needs a buyer vs. non-buyer split, and a surprising number don't have one. Without it, people who already bought keep getting "Finish your first order" discounts, which is a great way to train the best customers to sit on their hands and wait for a coupon they never needed. That comes straight out of margin. A structure pull shows in seconds which flows branch on purchase history and which are leaking discounts to full-price buyers.
- Account clutter: Old A/B drafts, abandoned builds, leftovers from a tool you tried 8 months ago: all of that piles up and skews the reporting we’re making decisions from. Pulling a census of the whole flow list surfaces what's stale so the numbers I'm trusting are actually clean.
The prompt I use for this: "Pull every live flow and show me the full structure: triggers, splits, delays, and the channel of each message. Walk through them and flag anything that looks like a build problem or a missed opportunity."
4. Catch deliverability decay before it costs you
Deliverability rarely fails loudly. It slips, an account starts landing in spam a little more, revenue softens with no campaign to blame, and weeks go by before anyone connects the dots.
What makes it easy to miss is that a healthy-looking average can sit right on top of a worsening trend: strong delivery and low spam complaints overall, while the most recent sends quietly drift the wrong way.
The only way to catch it is by looking at the trend send by send, in date order, which is tedious to do manually. MCP pulls every send chronologically in seconds and shows the line while there's still time to act.
The judgment part is knowing which warning to ignore. A win-back send to the most dormant addresses bouncing high is expected and self-contained. A run of healthy, everyday sends drifting up together is the real signal.
The prompt I use for this: "Show me delivery, bounce, spam-complaint, and unsubscribe rates for every send over the last 90 days, in date order. Is anything trending up, and which sends are the outliers?"
5. See the real list growth your net number is hiding
List growth is the number every brand watches, and it's the one that hides the most. A list can climb steadily on paper while swapping out most of its members every quarter.
The only way to see the truth is to pull adds and removals apart and look at the gross flow underneath, month by month, which is tedious enough manually that it rarely happens at all. That's exactly why churn stays invisible. MCP breaks it out in minutes, and what it surfaces often changes the whole plan.
The split is what matters. A text messaging list and an email segment can both show positive net growth, but the text messaging list keeps 3 of every 4 people it adds while the email segment loses 3 of every 4. It looks like the same direction on paper, but it’s completely different health underneath. One is compounding, while the other is a turnstile, with acquisition running flat out just to offset churn.
The prompt I use for this: "Show me my lists and key segments over the last 12 months: members added, removed, and net change per month. Where's the real growth, and where are adds just plugging churn?"
6. Turn complex analytics into weekly action
I'm far from the only one finding new ways to put MCP to work. Natalie Flowers is the director of digital marketing at Black Diamond Equipment, the outdoor gear brand that got their start making climbing equipment before expanding into hiking, trail running, and ski gear.
Flowers oversees acquisition and retention strategy there, and she's been using MCP with the Claude x Klaviyo integration to change how her team reports and reacts to data.
"We started connecting Claude to our Klaviyo data through MCP a few months ago, and it's already taken over our weekly reporting," Flowers says. "Every Monday, I give our leadership team a report on email, text messaging, and where our retention program stands overall. Now Claude builds that as an automated slide deck.”
“It pulls in our top campaigns and flows, documents them clearly, and goes deep on segmentation: which segments are performing best, which ones are dropping off, and sometimes that's because of deliverability issues on specific segments we wouldn't have caught as quickly otherwise,” Flowers adds.
It's not just the weekly report, either. "It also tracks our overall list growth every week and flags where the numbers are moving, so we can tell if a change we made to a flow or a pop-up actually helped,” Flowers explains
Flowers is candid about the fact that it took some iteration to get there. "The first few rounds needed a lot of feedback,” she admits. “The formatting needed work, the structure needed work. But the more we used it, the less feedback it needed. At this point I mostly just hand it the headline for the week, and some weeks I don't need to make any updates at all."
The time savings have been significant. "Conservatively, this is saving us 2–3 hours a week, and that's before you count the one-off asks,” Flowers estimates. “When someone on our leadership team has a random question, like how engagement looks across a certain segmentation, I used to have to pull the report and crunch the numbers myself. Now I just open a Claude chat and ask."
This saves us 2-3 hours a week. It takes out most of the digging and under-the-hood work.
For a brand that Flowers describes as slower than most to adopt AI, MCP has become a low-stakes way to prove out the value.
"We're honestly a bit behind on AI adoption as an organization, so this has been a good way to show leadership where it can actually be useful,” Flowers says. “Once MCP is set up as a connector, it shows up wherever we're working, whether that's Claude Chat, Cowork, or Code. We run it as a weekly automation that scrubs the account and hands us a list of what to act on for the following week. It takes out most of the digging and the under-the-hood work."
A quick note on where Klaviyo Composer fits in
MCP is one tool in my stack, not the whole stack. Lately I've also been digging into Klaviyo's built-in AI marketing agent, Composer.
MCP is something I run myself: I write the prompts, decide what to pull, and it only sees what Klaviyo's API exposes. Composer works differently. It lives right inside the campaign builder and runs on the full data layer behind Klaviyo, not just the API surface, so it's already working with real segment data, catalog data, and performance history instead of me feeding it in.
In practice, that split shows up in what I reach for each one to do. MCP is where I go when I have a specific analytical question and want to build the report myself. Composer is where I go for implementation: describe a campaign in plain language, like a spring reactivation for customers who've gone quiet for 90 days, on email and text messaging, in a specific tone, and it drafts the whole thing, audience, copy, and timing, ready for me to review.
Nothing goes live without a human approving it, and once brand voice and compliance rules are set up in Composer’s Knowledge Center, that structure carries through every output instead of me catching issues after the fact.
Composer was built to help marketers spend less time on production and more time on strategy. It’s not trying to replace the kind of analysis I do with MCP. Different jobs, same toolbox.
Related content
