June 4, 2026
How I’m using Kit MCP + Claude to get clear on my numbers
I have a decent open rate. Sitting right around 49% over the last 90 days across about 12,000 emails sent. That’s a number I’ve known for a while — and for a while, it was basically all I knew.
I knew the headline stat. I didn’t know what was underneath it.
So I asked. I connected my Kit account to Claude using the Kit MCP, pulled my last fifteen broadcasts, and asked Claude to tell me what it saw. What came back wasn’t a summary I wrote from memory — it was my actual data, live from my account, analyzed in the conversation.
Which emails were dragging the average down? Which topics were consistently landing versus consistently underperforming? Were my higher-click emails also the ones people were most likely to reply to, or was there a different profile for each? When did I last cover certain topics — and was the gap intentional or did I just drift away from them without noticing?
Those aren’t complicated questions. They’re just tedious to answer manually, which meant I mostly didn’t answer them. I’d glance at the open rate on a send, feel good or mildly annoyed depending on the number, and move on.
That changed when I started bringing real data into my AI conversations.
What “reviewing my numbers” actually looks like now
I use Kit as my email platform, and it has an MCP — a Model Context Protocol integration — that connects my Kit account directly to Claude. What that means in practice is that when I sit down to talk through my email strategy, Claude can actually pull my broadcast data in real time. Not a spreadsheet I exported. Not numbers I typed in myself. Live data from my account, available in the conversation as I’m working.
So instead of “here are my vibes about what’s been working,” I can ask specific questions and get specific answers.
Here’s an example of what I pulled recently. I asked Claude to pull my last fifteen broadcasts directly from Kit and show me what was performing above and below my baseline. Here’s what came back:
The sends that outperformed my baseline (49% open rate):
- “What I need from you to do some of my best work” — 54.5% open rate, 1.9% click rate
- “The two Airtable bases I’d start with to get your sh!t organized” — 54.2%, 1.66% clicks
- “How long have you been ‘figuring out’ Dubsado?!” — 50.8%
- “Notion vs Kitchen.co as a Client Portal” — 50.7% open rate, and my highest click rate at 3.1%
The ones that underperformed:
- “Acting brand new” — 21.5% open rate. That’s the outlier. That’s the one that tells you something.
- “Still acting like the new bish on the block” — 38.4%
That’s real data, pulled in real time from my Kit account during the same conversation where I was planning my next batch of emails.
Now I can actually ask: what do the top performers have in common? What’s different about the lower ones?
When I look at those numbers with Claude, the pattern that comes back isn’t complicated. My highest-performing emails tend to either name something specific my audience is already experiencing (the Dubsado one, the OBM one) or offer a direct comparison or resource (the Airtable bases, the Notion vs. Kitchen.co breakdown). The lower performers tend to be more abstract — a vibe or a mindset angle without a concrete anchor.
That’s genuinely useful. And it’s something I wouldn’t have seen just staring at a row of open rates.
The specific questions I ask
I’ve settled into a loose rhythm with this. Before I plan out a batch of emails or sit down to write, I’ll ask some version of these:
“What are my top 5 broadcasts by open rate in the last 60 days, and what do the subject lines have in common?”
This is the one that helps me understand what’s resonating before I go write more things. I’m not trying to reverse-engineer my voice into a formula — I’m just looking for signal.
“Which of my recent sends had the highest click rate, and what was the CTA?”
Open rate tells you the subject line worked. Click rate tells you the content and offer worked. Those are different things and they’re worth tracking separately. My Notion vs. Kitchen.co email had a 3.1% click rate — more than double my average — because it was a direct comparison with a clear link to act on. That’s worth knowing.
“When did I last send an email about [topic], and how did it perform?”
This one is mostly about avoiding gaps I don’t notice in real time. I write a lot about Airtable, ClickUp, and Dubsado. But there are whole areas — client onboarding process, proposal strategy, retainer structures — that I can let slide for weeks without realizing it. The data makes the gap visible.
“Which emails had lower-than-average open rates, and what was different about them?”
This is the uncomfortable one. But it’s the one that actually teaches you something. That 21.5% on “Acting brand new” is a flag. It’s not a reason to spiral, but it is a reason to look at the subject line format and see if something about it didn’t land.
What the Kit MCP makes possible
Here’s the thing about having this data available in a conversation: it changes the quality of the questions you can ask.
When I had to export to a spreadsheet or click through Kit’s individual broadcast reports, I’d answer one question and stop. The friction was just high enough that I’d get the number I came for and leave. Now I can ask a follow-up. And another. The conversation stays open long enough to actually learn something.
The Kit MCP gives Claude direct access to your broadcasts and stats, subscriber data, sequences, tags, and growth numbers — all without leaving the conversation. You connect it once and it’s there every time you work.
| Kit MCP + Claude | Manual export + AI | AI with no data | |
|---|---|---|---|
| Data freshness | Live, pulled in real time | As fresh as your last export | Not applicable |
| Time to get to insight | A few minutes | 15–30 min of setup per session | Immediate but based on guesses |
| What the analysis is based on | Your actual account history | Your actual account history | General best practices |
| Catch patterns across sends | Yes, automatically | Only if you think to look | No |
| Subject line strategy | Grounded in what’s worked for your list | Grounded in what’s worked for your list | Generic |
| Workflow friction | Minimal | Moderate | None, but output quality suffers |
The difference isn’t about having fancier AI. It’s about the AI having something real to work with.
If you’re on Kit and want to connect it to Claude, you can get set up here:
What I actually do with what I find
I want to be clear that I’m not running a full analytics review every week. This isn’t a 90-minute process. It’s closer to a ten-minute conversation I have before I plan a batch of emails, maybe once or twice a month.
What comes out of it is usually pretty simple. A few topic areas I’ve been avoiding. A subject line format I should use more. One or two emails from recent sends that I should probably build on because the engagement signals are telling me people want more of that conversation.
The goal isn’t to optimize myself into a predictable content machine. The goal is to not fly blind. I have a list that reads — a 49% average open rate is not an accident, and the people on it are telling me things through their behavior every time I send. Paying attention to that is just basic respect for the relationship.
The data was always there. I just needed a way to actually look at it without it becoming a whole project.
Using Kit as your email platform? The MCP integration is worth setting up if you want to bring your actual performance data into your AI workflows
This post contains affiliate links, which means we may receive compensation if you click on links to products or services and make a purchase. This comes at no additional cost to you
