Compare email MCPs by real launch work, connect Claude Code, and turn verified release evidence into a campaign your owned list can act on.
The best email MCP for Claude Code is the one that can complete your actual launch workflow with clear authentication, observable drafts, audience controls, and a deliberate delivery boundary. Lumail fits when you want Claude Code to operate campaigns, subscribers, workflows, and analytics while keeping consequential delivery behind an explicit approval path.
The goal is not to talk about an email platform. The goal is to ship a product change and let your owned list sell it while you keep building.
Do not choose an MCP by its tool count alone. Test one real release against these criteria:
| Criterion | Question to ask |
|---|---|
| Authentication | Is access tied to the correct account or organization, and can it be revoked? |
| Campaign state | Can the agent create a saved draft that a human can open and edit? |
| Audience evidence | Can it show the saved audience rule and recipient estimate? |
| Final rendering | Can it inspect the email with variables, links, and footer resolved? |
| Delivery boundary | Is scheduling or sending separated from drafting and review? |
| Reporting | Can it read the result back into the same release conversation? |
| Pricing fit | Does the billing model match list size and sending frequency? |
A polished writing assistant fails this test if it leaves you copying text between tools. A broad MCP can also fail if it hides the exact audience or makes delivery too easy to trigger accidentally.
As of August 14, 2026, this is no longer a category with one provider. Kit documents a 65-plus-tool MCP for paid plans, including Cursor and Claude clients. Lumail exposes an organization-aware tool catalog through MCP, with different safety profiles for OAuth plugins and API-token connections.
| Decision | Lumail | Kit |
|---|---|---|
| Best fit | Programmable campaigns, graph workflows, transactional email, and send-volume pricing | Creator newsletters, commerce, recommendations, and subscriber-tier pricing |
| Claude Code path | OAuth plugin with read and safe draft-write tools; full CLI beside it | Remote OAuth MCP on paid Kit plans |
| Delivery boundary | OAuth plugin does not expose send or schedule; use the app or CLI | Open-world actions return a Kit confirmation link |
| Pricing basis | Emails delivered; subscribers are unlimited on every plan | Active-subscriber tier on paid plans |
| Current catalog claim | 89 organization-scoped tools across the full catalog | 65-plus MCP tools in Kit's current help documentation |
Read the full Lumail versus Kit comparison and Kit's official MCP guide before buying. Features and prices change; the practical benchmark is one representative launch, not the table alone.
The Lumail plugin installs an OAuth MCP connection plus skills that teach Claude Code the workflow. Paste this into Claude Code:
Read lumail.io/claude-code/install and install the Lumail plugin for me.Claude Code runs the marketplace commands, opens the browser sign-in, and verifies the connection. You must approve the organization in the browser.
Open a new Claude Code conversation after installation. Claude Code captures its MCP tool list when a session starts, so the conversation that installed the plugin cannot use newly added tools.
For the manual commands, scope details, removal instructions, and troubleshooting, use the canonical Claude Code plugin guide. Anthropic's official MCP documentation explains the protocol from the client side.
Point it at the changelog, product page copy, approved price, screenshots, and the exact checkout URL. Tell it which source wins when facts conflict.
Use a bounded request:
Read the release notes and product-page copy in this repository. Create a launch campaign draft for subscribers tagged product-builders. Use only claims supported by those files, mark anything that needs live verification, and return the campaign ID, subject, preview text, saved audience rule, and recipient estimate. Do not schedule or send.Lead with the work the release removes or the result it unlocks. The stack can appear in one short proof section:
release evidence → Claude Code → Lumail draft → human review → owned list → offer
Lumail is the campaign layer. It should not replace the customer's problem, the product proof, or the offer in the narrative.
Ask Claude Code to read the campaign back. Verify the ID, subject, sender, audience, and every unsupported claim it found. Then open the same draft in Lumail's visual editor for final copy and layout changes.
Ask for the final HTML or text render with variables and the unsubscribe footer resolved. Check links against the production destinations. Repository URLs, preview deployments, and an HTTP 200 are not proof that checkout or onboarding works for a real buyer.
The OAuth plugin intentionally does not register send, schedule, publish, unsubscribe, or delete tools. Finish in the Lumail app, or use the full CLI in your terminal after an explicit send request:
npx lumail auth loginThis boundary lets Claude Code prepare nearly all the work without silently converting “draft this” into “contact my list.”
After delivery, ask for campaign delivery, clicks, and conversion events, then add the useful result to the next release brief. Do not treat an open rate as revenue. Compare the call-to-action click and purchase path with a similar audience and offer.
Claude Code can prepare and inspect a Lumail newsletter through the OAuth plugin, but that connection does not expose send or schedule tools. Final delivery happens in the Lumail app or through the Lumail CLI after explicit approval. This separation keeps the draft workflow agent-friendly while preserving a visible human decision before the list is contacted.
The seven-prompt pack already covers the recurring jobs behind this launch: turn a changelog into an issue, rank subject lines, segment clickers, review results, and build the follow-up sequence.
Get the seven prompts that run a newsletter, then use the Ship with Email hub for the complete operating loop.