Kit MCP Setup Guide: Connect Claude or ChatGPT to Your Email List (June 2026)

Dan Cumberland spent one day with Kit MCP and found 815 dormant subscribers he didn't know he had. He also found a newsletter that was quietly doubling as a booking funnel, two automations that had been broken long enough to matter, and more than 4,000 warm leads who had already clicked the link to his strategy call. His list is 12,000 subscribers on an AI consulting business, and most of that was sitting in plain sight inside his Kit account the whole time (Kit, June 11 2026). "Most of what I found in a day, I'd been sitting on for months," he told Kit. "The questions just weren't askable before."

That last line is the whole pitch. Until June 11, asking your email platform a question meant exporting a CSV, pasting it into a chat, and hoping the columns survived the trip. Kit MCP closes that loop. Connect Claude, ChatGPT, Cursor, or Gemini once, and your AI tool reads your real Kit data and acts on it in the same conversation.

What changed on June 11

Kit (the platform formerly called ConvertKit) shipped nine product launches at its Craft + Commerce conference in Boise. Most were the usual feature-list fare. The one that matters for anyone running a list with an AI assistant open in the next tab: Kit MCP came out of public beta, now available on the Creator and Pro plans (Kit).

MCP stands for Model Context Protocol, the open standard that lets an AI tool talk to an outside service through a defined set of actions. Plenty of platforms ship one. Most are read-only: the AI can look at your data and report back, but it can't do anything. Kit went the other way. Per its own announcement, the Kit MCP exposes 76 endpoints and reads and writes, which Kit calls "the most complete email MCP that exists." Over 3,000 creators were already using it before it left beta (Kit).

What Kit MCP actually is

Strip away the marketing and it's a bridge. On one side, your AI tool. On the other, the same Kit account you log into. The 76 endpoints are the verbs in between: pull a segment, score subscriber engagement, draft a broadcast, apply a tag, trigger an automation, check campaign stats. The Kit feature page frames the workflow in three moves: connect your tool, ask questions about your list, then put the assistant to work acting on what it found (Kit MCP).

The read-and-write part is the difference that's easy to undersell. A read-only MCP can tell you that 815 people haven't opened anything in six months. A read-and-write MCP can tag those 815, draft the win-back email, and queue it, all without you touching the Kit UI. Kit's own description: read and write means "not only analyzing the data in your Kit account, but also immediately taking action on it directly from your AI tool" (Kit MCP).

You stay in the loop on the writes. Every action that changes something waits for your approval before it fires. You set the permissions up front and can revoke access anytime (Kit MCP). More on the gotchas there later.

Which plan you need

Here's the part the announcement buries. The free Newsletter plan does not include MCP. You need Creator or Pro. Current pricing pulled from Kit's page on June 13:

Plan	Price (Kit pricing)	Subscriber cap	Kit MCP	Best for
Newsletter	$0/mo	Up to 10,000	No	A new list with no AI workflow yet
Creator	$33/mo ($390/yr)	1,000 base, scales with list size	Yes	Solo operators who want AI in the loop
Pro	$66/mo ($790/yr)	1,000 base, scales with list size	Yes	Creators who also want Subscriber Signals + Engagement Analytics

Pricing per Kit's pricing page, fetched 2026-06-13. The subscriber cap on Creator and Pro scales with how many people are on your list; the dollar figures above are the entry tier. The free Newsletter plan covers up to 10,000 subscribers, which is generous, but it's a hard wall for MCP. If your reason for being here is the AI workflow, the free tier doesn't get you in the door.

The clear read (Kit pricing): for most operators, Creator at $33/mo is the pick. It's the cheapest tier with MCP, and the 76 endpoints don't change between Creator and Pro. What Pro buys you on top is the data layer, Subscriber Signals and Engagement Analytics, which make the MCP smarter because there's more for it to read. We'll come back to who needs that.

You can compare the tiers and start on Kit directly.

Setup in five steps

This is a connector setup, not a code install. No CLI, no API keys to rotate by hand. The flow Kit documents is short (Kit MCP):

Upgrade to Creator or Pro. MCP is gated behind these plans. If you're on the free Newsletter tier, this is the only blocking step.
Grab Kit's MCP server URL. It lives in your Kit account's AI/MCP settings. This is the address your AI tool will point at.
Paste it into your AI tool's connector settings. In Claude, that's the Connectors panel. In ChatGPT, the connector/MCP settings. Cursor and Gemini have their own equivalents. The pattern is the same: a "remote MCP server" field where you drop the URL.
```
Tool:        Claude (Settings > Connectors > Add custom connector)
Field:       Remote MCP server URL
Value:       <your-kit-mcp-server-url>   # from Kit AI/MCP settings
Auth:        OAuth handshake, approve in the Kit tab that opens
```
Set permissions. This is the step people skip and regret. Decide what the AI tool is allowed to do before you give it a real task. Start read-only if you're nervous; you can widen scope later. You can revoke the whole connection from Kit at any time.
Run a read-only query first. Don't open with "delete the dead subscribers." Open with a question. Something like:
```
What's my average open rate over the last 90 days
compared to the 90 days before that?
```
Kit's own example screen shows exactly this prompt returning a before/after table (Kit MCP). If the numbers match what you see in your Kit dashboard, the connection is live and you can trust it. Then graduate to writes.

The "read-only first" habit is the one worth keeping. An MCP write is real. When the assistant says it tagged 815 people, it tagged 815 people. Confirm the read path is accurate before you hand it the keys to the write path.

Three things Kit MCP does that a manual workflow can't

Plenty of this you could do by hand. The point isn't that it's possible, it's that the friction drops to near zero, which changes which questions you bother asking.

A dormant-subscriber audit in one prompt. The manual version: export your list, sort by last-open date, define "dormant," count, re-import a tag. Twenty minutes if you know your way around a spreadsheet. The MCP version is a sentence: "Find everyone who hasn't opened an email in 180 days and tag them dormant." This is the exact shape of what Dan Cumberland surfaced. The 815 dormant subscribers weren't hidden; they were just expensive to ask about (Kit). MCP makes the question cheap, so you actually ask it.

A segment built from a conversation, not a query builder. Kit's visual segment editor is fine. It's also a series of dropdowns. With MCP you describe the segment in plain language, the AI translates it into the underlying conditions, and you approve the result. "Subscribers who clicked the launch email but didn't buy, and who joined in the last 90 days" becomes a tagged segment without you mapping a single filter rule.

A broadcast drafted from your own campaign data. This is where read-and-write earns its keep. You can ask the assistant to look at what your best-performing recent broadcasts had in common, draft the next one in that shape, and create it in Kit as a draft for you to review. Kit's product screen shows precisely this: a prompt requesting a broadcast, an "Approve & create broadcast" button, and a drafted broadcast waiting in Kit (Kit MCP). The draft never sends on its own. You approve, you edit, you hit send.

Want the workflow to spill outside Kit? Because Kit MCP can fire automations, you can use a Kit automation as the trigger for a downstream tool like Make, so an AI-tagged segment in Kit kicks off a multi-step scenario you've already built. That's an extension, not a core feature, but it's the natural next step once the MCP is wired in.

Where Kit MCP falls short today

It's new, and new shows. A few candid limits.

It is not a payments tool. The 76 endpoints cover list operations: subscribers, tags, segments, broadcasts, automations, reporting. They don't hand the AI control of your Kit Commerce checkout or your payouts, which is the right call for a first release but worth knowing if you imagined "run my whole store from Claude."

It doesn't build automations from scratch through the AI. Kit MCP can trigger automations you've already built. It is not a visual automation builder you drive by prompt. The complex branching logic still gets assembled in Kit's automation editor; the MCP fires it.

The write actions need babysitting at first. The approval gate is a feature, but it also means you can't fully fire-and-forget a destructive change until you trust the tool's reads on your specific account. Budget a session of read-only verification before you let it tag thousands of people.

And the data layer that makes the MCP genuinely smart sits on Pro. The richest reads, who your highest-value subscribers actually are, come from Subscriber Signals, which enriches your list with job title, income, location, and influence markers (Kit). That's Pro-only. On Creator you get the full 76-endpoint MCP, but the AI is reasoning over engagement and tags, not enrichment data. Tarzan Kay used Signals to tag 1,066 subscribers as potential affiliate partners in a single launch (Kit); that specific move needs the Pro data, not just the MCP.

Kit vs Beehiiv vs Substack for AI-first creators

If you're choosing a platform in 2026 specifically because you want an AI assistant wired into your list, the MCP question reshapes the usual comparison. Kit is the only one of the three shipping a read-and-write MCP with this depth right now: 76 endpoints, both reads and writes, on a $33/mo plan (Kit). Beehiiv is the stronger platform if your priority is paid-newsletter growth mechanics and its own native AI writing and analytics features, and it remains the closest competitor on monetization shape. Substack is the simplest to start and the best distribution network for a writer who wants discovery over automation, but it gives you the least programmatic control of your list, which is exactly the lever an MCP pulls. The broader platform call still comes down to monetization shape and where your readers already are. The narrower call for this article is cleaner. If "Claude should be able to act on my list" is a real requirement, Kit is the one of the three that does it today.

The bottom line

Wire up Kit MCP if you already run an AI tool next to your newsletter and you've ever exported a CSV to ask it a question. That's the whole tell. The export-then-paste habit is the friction Kit MCP removes, and the payoff is that you start asking questions you used to skip.

For a solo operator, go Creator at $33/mo (Kit pricing). You get all 76 endpoints, the read-and-write workflow, and the cheapest door into the feature. Run the dormant-subscriber audit in your first session and the upgrade pays for itself in re-engaged readers.

Upgrade to Pro at $66/mo (Kit pricing) if your list is your business and you sell to it. Pro adds Subscriber Signals and Engagement Analytics, which give the MCP richer data to reason over: who your high-value subscribers are, which acquisition source brought the readers who stuck, what a subscriber relationship is worth over time. If you're matching sponsors to your audience or recruiting affiliate partners the way Tarzan Kay did, that data layer is the difference, and the MCP is how you act on it without leaving your chat.

Stay on the free Newsletter plan only if you're not ready to use the AI workflow at all. It covers 10,000 subscribers for free, which is a genuinely good deal for a new list, but it does not include MCP. The moment "I want Claude to read my list" becomes true, you're on Creator.

You can start or upgrade on Kit and have your first read-only query running in the time it takes to paste one URL.