Testing Win-Back Subject Lines in Cursor
Dormant subscribers need a sharper hook than your weekly newsletter. This guide shows how to draft, test, and pick win-back subject lines directly inside Cursor by connecting marginal, a hosted email MCP server that gives your AI agent real send-and-measure tools.
Why win-back subject lines deserve their own test
Win-back emails fight an uphill battle: the recipient stopped engaging for a reason, and the subject line is your only chance to interrupt that habit. A subject that works for active subscribers — short, casual, on-brand — often underperforms against a re-engagement audience that needs a reason to care again.
Because the stakes and the psychology differ, win-back deserves a dedicated subject line A/B test rather than reused copy. Testing inside Cursor lets you keep the loop tight: iterate on angles, launch, and read results without switching to a separate marketing dashboard.
- Curiosity vs. incentive: "We saved your spot" against "Here's 20% to come back"
- Loss framing: "Your account is about to go quiet"
- Personal check-in: "Did we do something wrong?"
- Direct value reminder: "Three things you've missed"
Connect Cursor to the marginal email MCP server
marginal exposes four MCP tools — generate_variants, launch_test, get_results, and recommend_next — over the endpoint at https://marginal.sh/mcp. Once Cursor can reach those tools, your AI agent can run an end-to-end win-back subject line test on your behalf.
Add marginal as an MCP server in Cursor's configuration and authenticate with your Bearer API key (it starts with marg_live_). Because marginal is fully hosted, there's nothing to deploy or self-host — point Cursor at the endpoint and you're ready to test.
- Registry id: sh.marginal/mcp
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Setup reference: https://marginal.sh/mcp
Run the win-back test from your editor
Describe the campaign to your Cursor agent in plain language — "win-back for subscribers inactive 90+ days, four subject line variants" — and let it call the tools in sequence. generate_variants drafts the candidate lines, launch_test schedules the managed send and splits the audience, and get_results reports open and click data with within-test lift metrics.
When the data is in, recommend_next suggests the winning variant and a sensible follow-up, so you can move from a single subject line A/B test to an ongoing re-engagement cadence without leaving Cursor.
- generate_variants — produce win-back subject line candidates
- launch_test — split the dormant segment and send
- get_results — open/click tracking plus within-test lift
- recommend_next — pick the winner and plan the follow-up
Tips for an AI agent email marketing workflow
Keeping subject line tests inside Cursor pairs naturally with the way developers and growth engineers already work — version your campaign prompts, reuse them across segments, and let the agent handle the mechanical send-and-measure steps.
- Hold the audience and offer constant so the subject line is the only variable
- Test contrasting angles, not minor word swaps, on a re-engagement list
- Give each variant enough volume before reading lift
- The free tier covers 100 experiments/month — plenty for iterating on win-back copy
- Full tool reference lives at https://marginal.sh/docs/
Get started with marginal
marginal is a hosted email marketing MCP server at marginal.sh. Sign up free, create an API key, and connect https://marginal.sh/mcp from Cursor, Codex, or Claude Desktop.
- 100 experiments/month on the free tier
- Four MCP tools: generate_variants, launch_test, get_results, recommend_next
- Listed in the MCP Registry as sh.marginal/mcp