Most creators are using AI to write emails. Operators use AI to run the system. Kit just shipped the bridge.
Until last week, AI lived next to your email tool. You drafted in Claude. You copy-pasted into Kit. Copy-paste was the bottleneck.
MCP collapses that. Claude doesn't draft an email for you to paste into Kit. Claude opens Kit, reads your last 12 broadcasts, identifies the three subject lines that beat your average, drafts the next one in that pattern, and creates the draft for you to review. One conversation. No tab-switching.
This is a different relationship with your tools. Less clicking. More thinking. The creators who treat their email system as a programmable surface, not a clickable dashboard, are going to run quietly past the ones still copy-pasting subject lines from a Google Doc.
Most "AI for email" coverage focuses on copy generation. That's the small win. The big win is operations. Audits, segmentation, list health, re-engagement, sequence editing. The work nobody wants to do that quietly determines whether your business compounds.
This guide is the operator's playbook for that work. 50 prompts. 10 workflows. Setup takes 4 minutes. Then we'll walk through what you can actually do.
The MCP is paid-plan-only. The free Newsletter plan does not include MCP access. Creator runs $33 per month at 1,000 subscribers (or $390 per year, saving $78). Pro runs $66 per month at the same level (or $790 per year, saving $158). 14-day free trial, no credit card required.
Start here. Get your Kit account set up before you do anything else. The MCP only works once you're on Creator or above.
Start your Kit account →The Kit MCP URL is https://app.kit.com/kit-mcp
Compatible clients: Claude.ai, Claude Desktop, ChatGPT, Cursor, Windsurf, Microsoft Copilot, Claude Code.
If you're on Claude.ai (browser): Go to claude.ai/settings/connectors, add a custom connector, paste the MCP URL, and authorize through Kit.
If you're on Claude Desktop: Open Settings, find the Connectors section, paste the MCP URL, and authenticate. A browser tab opens. Select the Kit account, approve access. Tab closes. You're connected. Fully quit and reopen the app once.
Kit's MCP defaults to per-call approval. The first time Claude runs a Kit tool, you'll see an approval dialog. This is a feature, not a bug. Three rules for how to use it:
Type this into Claude:
You'll see the approval dialog. Approve it. Claude returns new subscribers, cancellations, and net growth from your live account. The MCP is working.
Kit's docs say "50+." The actual count is 65, across 17 categories. Here's the operator's view: what each category lets you do, and which risk tier the tools sit in.
Pull email stats, growth stats, identity, brand colors, creator profile. Update brand colors.
List, get, create, update single or bulk. Filter by engagement and sign-up date. Pull per-subscriber engagement stats. Unsubscribe.
List, create, rename. Tag and untag single or bulk (up to 100 sync, larger batches async).
List, create, rename, delete (destructive). Bulk update values across many subscribers.
List segments, get details, get the segment DSL schema for advanced filtering.
List forms and landing pages, list subscribers per form, add subscriber to a form.
List, get, create, update, delete (destructive). Add subscribers to sequences.
List, get, create, update, delete individual emails within a sequence. The killer differentiator.
List, get, create draft, update, delete. Per-broadcast stats, account-wide stats, per-link clicks. The MCP never sends. You approve from Kit's UI.
List published posts, get a specific post with full HTML.
List email templates. List, get, create, update reusable content blocks.
List, create, delete webhooks for real-time event delivery to your tools.
List authors, condition sets, content tags. Plus the meta-tool: list_prompt_suggestions.
Read Safe. Pulls data. Never changes account state. Around 38 of the 65 tools sit here. "Always allow" these and stop worrying.
Write Reversible. Tags, drafts, updates. Around 22 tools. Approve case-by-case until you trust the pattern, then "always allow" the ones you use weekly.
Destructive Irreversible. Around 5 tools: delete broadcast, delete sequence, delete custom field, delete webhook, unsubscribe. Keep manual approval on these forever.
Most people approach AI tools like a vending machine. Insert prompt, receive output. The operators who get real value treat it like a junior CMO who happens to have read every marketing book ever written and can pull live data from your tools.
Ask Claude to pull data and propose options. You decide. "Find my top 5 subject lines and propose 3 patterns I could replicate" beats "Write me a subject line."
Big prompts that try to do five things at once underperform. Walk Claude through the work: pull this, analyze that, draft based on the analysis. The chain becomes the workflow. The workflow becomes the operating system.
Ask Claude to cite specific broadcast IDs, subscriber counts, click-throughs. Numbers ground the work and let you verify. "Show me my last 5 broadcasts ranked by open rate, with subject lines and exact open percentages" beats "Tell me which subjects worked."
Most prompt libraries organize by tool. That's useful for engineers. Operators don't think in tools. They think in jobs to be done. Tap any prompt to copy.
A note on data: These prompts work on day 1 or year 3. Claude will tell you when the data is thin. Cross-tool prompts assume you have the relevant MCP connected.
These assume you have the relevant MCP connected. Stripe and Meta have public MCPs. ThriveCart-style commerce MCPs are coming.
Single prompts are nice. Workflows are where the leverage lives. Each of these is a chain of prompts that produces a real operating outcome.
Read Pull the week's broadcast stats, identify winners and losers, capture lessons for next week.
ReadWrite The gateway use case. Draft this week's newsletter with your voice and your data.
Read Find what's actually working in your subject lines so you stop guessing.
ReadWrite The maintenance pass that keeps deliverability healthy.
ReadWrite Build a 3-email re-engagement sequence from inactive subscriber data.
ReadWriteDestructive Audit every tag, kill duplicates, fix orphans. Approve destructive actions manually.
ReadWrite Sync your paying customers across Kit and your commerce stack so segmentation stays clean.
Read The strategic step-back. Generate a one-page CMO-style report for your own business.
ReadWrite Capture an idea the moment you have it. Turn a 3-minute voice memo into a drafted broadcast before the thought fades.
Removes the "I'll write it later" barrier. The moment the idea shows up is when it's sharpest.
Read Every creator with a long sending history is sitting on a content library they've never had time to organize. This workflow builds it once and keeps it current.
Two to four years of broadcasts you've forgotten about is a content goldmine. This workflow unlocks it.
Run these every quarter and the system runs without you. Workflows are the part most creators skip. They're also the part that compounds.
Try Kit and run them yourself →Bonus workflow: Newsletter Sponsor Reporting. If you run sponsored sends, build a recurring workflow to pull broadcast stats and sponsor-specific click data, then draft a sponsor report email. Same shape as the others. Different audience.
Marketing AI Playbook is the operator's library for the AI era of marketing. Real prompts, real workflows, no fluff. Grab the Starter Pack and the next operator's guide will land in your inbox the moment it ships.
Get the Starter Pack →An MCP on its own is useful. Multiple MCPs in the same conversation is where the leverage compounds. Here are the combinations worth thinking about.
Pull revenue data. Identify paying customers. Tag them in Kit. Build a "high-LTV" segment that crosses purchase value, engagement, and tenure. Run monthly to keep the segment fresh.
Sync Custom Audiences from Kit segments to Meta. Build lookalike audiences from your most engaged subscribers. Run a Meta campaign targeting an audience that mirrors your top decile. Pull the campaign results back into Claude to assess.
Pull research from a Google Doc and drop it into a broadcast draft. Save broadcast reports to a Drive folder. Read meeting notes and draft a relevant newsletter. The unsexy combo that saves hours weekly.
For B2B operators, bidirectional sync of subscriber-as-lead. Find Kit subscribers who match your CRM's ICP criteria. Tag them for VIP nurture. Pull CRM stage data back to Kit for segmentation.
When ThriveCart, Lemon Squeezy, or other commerce platforms ship MCPs, the customer sync workflow gets much cleaner. For now, do it manually monthly with the data you can pull. The pattern is the same.
Both Kit and Beehiiv shipped MCPs. They're not the same. Here's the head-to-head from someone who built operator's guides for both.
| Dimension | Kit MCP | Beehiiv MCP | Edge |
|---|---|---|---|
| Total tools | 65 | 56 | Kit |
| Approval model | Per-call by default | Per-session | Kit (safer), Beehiiv (faster) |
| Send and schedule via AI | No, returns confirm_url | Yes, full scheduling | Beehiiv for speed, Kit for safety |
| Engagement filtering | Atomic, AND-able, scoped to broadcasts and URLs | Segment DSL (SQL-like) | Kit for AI-driven assembly, Beehiiv for power users |
| Sequence email editing | Full read and write | Read only | Kit |
| Built-in prompt catalog | Yes (list_prompt_suggestions) | No | Kit |
| Bulk operations | First-class, async over 100 | Yes | Tie |
| Cross-tool composition | Strong via standard MCP patterns | Strong via standard MCP patterns | Tie |
| Documentation | Notion hub plus help docs | Developer portal plus help docs | Tie |
| Pricing | Creator plan and above ($33/mo) | Most paid plans | Tie |
Beehiiv's MCP is faster to ship a broadcast end-to-end. Kit's is more architecturally cautious and more powerful for ongoing list operations. If your business is built on automated sequences, Kit's deeper sequence-email write access matters. If your business is broadcast-heavy and you trust AI to send, Beehiiv's send-tools save real time. Both are good. They're optimizing for different operators.
For most solo marketers and small teams, the deciding factor isn't the MCP. It's the underlying platform: deliverability, automation depth, monetization features, community. The MCP is the multiplier. Pick the platform that fits your business, then use the MCP to run it.
If you're choosing between them or migrating, this is the link.
Start with Kit →Three operational tasks every Kit user runs. Same job, before and after MCP.
Open Kit. Click Subscribers. Filter by tag. Adjust date range. Export to CSV. Open in Sheets. Sort by activity. Look for patterns. Switch back to Kit. Apply tags manually.
"Find subscribers tagged X who clicked URL Y in the last 30 days, and tag them Z." Done in 20 seconds.
Click through 20 broadcasts manually. Note the open rate for each in a spreadsheet. Sort. Try to remember which subject lines worked. Lose track. Start over.
"Rank my last 90 days of broadcasts by open rate. Show me subject lines and exact percentages." 15 seconds.
Open the sequence. Click each email. Read it. Try to remember what the previous one said. Wonder which one has the lowest open rate. Click into stats. Repeat.
"Audit my onboarding sequence. Identify the email with the lowest open rate. Propose three rewrites in my voice." 90 seconds.
Dan Cumberland runs a 12,000-subscriber list and had built his own Python analytics scripts and browser automation just to get data Kit's UI couldn't surface. Then he spent one day with the Kit MCP. What he found is worth reading before you run your first workflow.
Dan's story was documented by Cait Miller on the Kit blog. The full case study is here. What follows is the operator's summary.
Dan didn't come in looking to replace what he already had. He came in to find what was previously unaskable. He found five things.
Kit's segment builder can't combine "opened 30+ times historically" with "zero opens since March." The MCP can. Dan ran three cohorts with tightening thresholds and landed on 815 subscribers who had been deeply engaged and then went completely dark. That's a re-engagement sequence with a warm lead pool that didn't exist before he asked the question.
Dan sells consulting to architecture, engineering, and construction firms. He'd wanted a hard ICP penetration number for years. The answer had always required a CSV export and three to four hours of manual classification. With the MCP, he sampled 600 subscribers across six paginated calls, cross-referenced industry field, title keywords, and company name — and had his number in 10 minutes. About 4.5% of his list matched his ICP. That's roughly 570 people, not 12,000. That one number changed how he thinks about list-building.
Dan pulled his full purchases list, sampled high-value buyers, and traced each one back to their original opt-in form using Kit's kit_first_form custom field. One $2,000 cohort buyer traced back to a specific lead magnet he might have otherwise deprioritized. Before the MCP, that attribution pass took roughly 10 minutes per buyer through the Kit UI. Across a quarterly review, that's an eight-hour project nobody runs. With the MCP, it's 30 minutes.
Three parallel MCP calls — sequences, forms, tags — plus a walk of the exclusion rules on each sequence, rendered as a Mermaid flowchart. Fifteen minutes. Dan's diagram revealed a four-way role-routing split he'd built over time but had never seen mapped end-to-end. It also surfaced roughly 25 retired sequences from a previous brand era still marked active in Kit. An automation graveyard, invisible until someone went looking.
Dan pulled his 30 most recent broadcasts with open rates, hand-derived a scoring rubric from his top performers, and generated new subject line variants scored against that rubric. The top scorer was "I stopped selling AI tools to engineering firms" — matching the first-person confession pattern his list historically responds to. What had been 20 to 30 minutes of inconclusive brainstorming per send is now a three-minute exercise, trained on his actual data, not generic best practices.
Dan had built Python scripts and browser automation just to ask better questions about his own list. The MCP didn't replace his rigor. It made his questions answerable without the scaffolding. If a power user found eight things he hadn't known about a list he'd managed for years, the question isn't whether your list has similar gaps. It's which ones you'll find first.
Kit shipped 65 tools on day one. They've signaled more are coming. The MCP isn't the destination. It's the new starting line.
Three things to expect over the next 12 months:
Kit will keep expanding the surface area. Watch for tools that close the remaining gaps: automation creation, sequence email deletion, deeper template control.
Every major SaaS is shipping or considering an MCP. The operators who plan for cross-MCP composition will run circles around the ones still working tool-by-tool.
The creators who treat email as a programmable surface, not a clickable dashboard, will compound faster than the ones still copy-pasting subject lines. Not because they're working harder. Because the system is doing the work.
This is where it starts. Kit Creator or above. Connect the MCP. Run a few workflows. Watch what changes.
Build it on Kit →This guide is part of Marketing AI Playbook, an operator's library of MCP and AI tool guides for solo marketers and small teams. Built and maintained by Tuck Ross, former Fortune 500 marketing executive.
If you made it this far, you're the kind of operator Marketing AI Playbook is built for. Grab the Starter Pack and the next operator's guide ships to your inbox the moment it's live. Real prompts, real workflows, no fluff.
Get the Starter Pack →