Playbook · Mail
Convert cold email signups by landing in Gmail Primary.
Someone signed up. They went quiet. A 3-email welcome sequence written to land in Gmail's main inbox — not Promotions — is the cheapest re-engagement tool you have. Setup takes about ten minutes; every new signup gets the sequence automatically after that.
https://loopi.social/playbooks/land-in-gmail-primary — paste it into Claude Code or Codex with Loopi connected, and say "Read this Loopi playbook and run it for me." The agent fetches the page and builds the whole sequence — emails drafted, previewed to your inbox, and scheduled — without you writing a line of copy.What it does for you
Most cold signups go quiet before you ever have a real conversation with them. This play fixes that: a 3-email sequence that arrives in Gmail's main inbox (not the Promotions tab, where people don't read), written in your voice, warming each signup toward a real decision. By the end you'll have:
- A 3-email welcome sequence drafted, previewed, and live. Your AI agent writes all three emails, sends a test copy to your own inbox so you can confirm the tab before any subscriber sees it, then schedules the sequence automatically.
- Each email pointing at a different section of your landing page. Email 1 introduces the problem. Email 2 gives proof. Email 3 makes the soft ask. Every email ends with one CTA back to wherever that conversation belongs.
- A signup form snippet ready to paste into your site. Anyone who fills it out goes straight into the sequence — no list management, no Zapier.
- Primary inbox placement from the first send. Gmail classifies every message on its own. The agent follows the copy rules that keep messages out of Promotions — first-person voice, concrete proof, one ask — and re-previews if anything slips.
How it works, start to finish
Three steps, two of which are one-time setup. Here's the whole thing, nothing hidden.
1 · Loopi account
60 seconds — email + password
2 · Connect MCP
One line wires Loopi to your agent
3 · Run the prompt
Agent drafts, previews, schedules
1. Create a Loopi account
The prerequisite is a Loopi account — 60 seconds at loopi.social (email + password). Inside the app, the Mail tab is where your lists, sequences, and content templates live. You don't need to configure anything before running the prompt — the agent sets up the list and sequence in the same run.
2. Connect Loopi to your agent
Loopi is an MCP server. One line wires it to Claude Code. Claude Desktop, Cursor, Codex, and Gemini take the same https://api.loopi.social/mcp URL in their connector settings. Authorizing opens a browser once — after that, the connection stays.
3. Run the prompt
Paste this into Claude Code (or any connected agent). Swap in your real landing page URL. The agent handles everything else — it drafts all three emails, sends previews to your inbox, fixes anything that slips into Promotions, then schedules the sequence and returns the signup form snippet.
Wire this play up — connect Loopi to your AI agent →
What the agent does under the hood
You hand the agent three things. It returns three.
You provide
You get back
Landing page URL
So the agent knows what it's selling
Your voice brief
First-person, one proof anchor per email
Loopi MCP connected
The agent calls the tools directly
AI agent
Claude, GPT, etc. via Loopi MCP
3-email sequence
Drafted, previewed, scheduled
Signup form snippet
Paste into landing page
Live on your list
New signups get email 1 within minutes
What Loopi handles for you
Net effect: the only knobs you touch are landing page URL, voice, and tone. The plumbing happens behind the API.
- Compliance, automatically. Loopi auto-injects a compliant unsubscribe footer on every send — you never write one manually. This also matters for inbox placement: a bordered "Sent via X" branding block tips Promotions by itself, so Loopi keeps the footer minimal and inline.
- Preview before any subscriber sees it.
mail.previewContentsends a real copy of the email to your inbox. You open Gmail, check the tab, confirm Primary before the agent schedules anything. If it slips, the agent diagnoses the body and re-previews. - Click tracking on every link. All links are rewritten through Loopi's click tracker — so when you check analytics, you can see which email drove the most landing page visits.
- Drip scheduling and sequencing.
mail.createSequencewires the three emails into a timed sequence. New subscribers who fill out the form enter the sequence automatically — no manual scheduling, no re-running the prompt. - The signup form, one call.
mail.createFormreturns a paste-ready embed snippet — a single tag you drop into any site (plain HTML, WordPress, React, Astro). Submissions go straight into the list.
The evidence: what actually tips Primary
We sent roughly 100 controlled variants to the same Gmail inbox over a few weeks. Held everything constant except one variable per send — body tone, footer style, subject framing, threading headers, HTML chrome, sending infrastructure. Watched which tab each landed in.
Two findings dominated everything else:
- Body phrasing is the single biggest factor. Conversational, first-person, addressing one reader lands Primary. Declarative product-pitch tone — "Only in 2026 can a single person ship…" — tips Promotions every time, even with stripped HTML and no footer.
- A bordered "Sent via X" footer tips Promotions by itself, even on an otherwise-Primary body. Loopi's auto-injected footer is minimal and inline — it doesn't trigger the classifier.
Things that didn't matter the way we expected: styled CTA buttons (safe), threading headers (UX-only, no classifier effect), sending platform (SES with the right body lands Primary as cleanly as Gmail API).
Side-by-side: same product, opposite outcomes
Both bodies pitch the same course. Same conversational shape, same opener, same speaker. The left tipped Promo; the right landed Primary. The difference is one paragraph sliding from "telling my story" into "describing the product."
✗ Promo — product description
✓ Primary — speaker's story
The diagnostic line: talk about what you built or noticed (an app, an experience, a year of engineering), not what the product offers (missions, features, pacing). The moment a sentence starts pitching the product's structure, the classifier flags it.
Body anti-patterns that tip Promo
- "Now live", "is here", "is now available", "limited time", "X% off" — launch keywords.
- "Real product, real users, real money" — slogan cadence.
- "The window is open, most people won't take it" — FOMO.
- "Bring your own idea. Follow the missions. Ship your MVP in 7 days" — imperative slogan close.
- "From the team at X", "your friends at X" — brand-team voice.
- "X has Y built in" — feature-pitch statement. Rephrase as a verb of doing: "I baked Y into X".
- "When your agent does X, ask it to Y" — second-person tutorial cadence. Rephrase first-person: "When I ask my agent for X, what comes out is Y".
Pricing is fine. Sale-framing is not.
- ✓ "The first batch gets it at $97 — I priced it lower so I can stay close to everyone going through it. After that, it's $297."
- ✗ "Inside the cap: $97. After: $297. No second window." — fragment-sentence cadence + FOMO close.
Ship this play from your AI agent.
Connect Loopi via MCP, paste the prompt, and your agent drafts the sequence, previews every email to your inbox, and schedules it — in one run.
Create an account →