← All guides marginal.sh

How to Test Urgency vs Curiosity Subject Lines

Updated June 2026 · Hosted email MCP server

Urgency-driven subject lines ('Last 6 hours to save') and curiosity-driven ones ('You left something behind') pull on different psychological levers — and the winner is rarely the one you'd guess. This how-to guide walks through testing urgency vs curiosity subject lines directly from your AI agent using marginal, the hosted email MCP server.

Frame the test before you generate variants

Before generating subject lines, decide what each style is supposed to do. Urgency leans on scarcity, deadlines, and FOMO. Curiosity withholds information to spark a click. A clean subject line A/B test isolates one variable, so keep the offer, audience, and send time identical across both arms.

Generate urgency and curiosity variants

Connect your agent — Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, or OpenAI Codex — to the marginal MCP server at https://marginal.sh/mcp using your Bearer API key. Then call generate_variants and ask explicitly for the two contrasting styles.

Prompt your agent with something like: 'Generate 3 urgency subject lines and 3 curiosity subject lines for our cart-abandonment email.' That gives generate_variants a balanced set so each style is fairly represented in the test.

Launch the test and let marginal manage the send

Use launch_test to send the urgency and curiosity arms to your audience. marginal handles the managed send and tracks opens and clicks, so you don't wire up your own delivery or tracking pipeline. The free tier covers 100 experiments per month, which is plenty for iterating on subject line styles.

Read the results and decide your next move

Call get_results to see within-test lift between the urgency and curiosity arms. Lift metrics compare each variant against the others in the same test, so you get an apples-to-apples read on which psychological lever actually won for this audience.

Then ask your AI agent to call recommend_next. It uses the test outcome to suggest the next experiment — for example, doubling down on the winning style with a sharper hook, or testing whether curiosity holds up against a stronger deadline. This is the loop that makes AI agent email marketing compounding rather than one-off. See the marginal MCP tutorial in the docs at https://marginal.sh/docs/ to go deeper.

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.

Documentation · Sign up · FAQ