Testing Upsell Subject Lines in Continue
Continue brings an AI coding assistant into your editor — and with marginal's email MCP server, that same assistant can draft and run upsell subject line A/B tests. This guide shows how to wire the two together and turn upgrade nudges into measured experiments.
Connect marginal to Continue
Continue supports MCP servers, so you can expose marginal's tools directly to the model you chat with inside your editor. marginal is hosted at https://marginal.sh/mcp, which means there's no self-hosted process to babysit — you just point Continue at the endpoint and authenticate with a Bearer API key.
Add the server to your Continue MCP configuration using your `marg_live_...` key. Once connected, the model can call generate_variants, launch_test, get_results, and recommend_next as part of a normal conversation.
- Endpoint: https://marginal.sh/mcp
- Registry id: sh.marginal/mcp
- Auth: Bearer API key (marg_live_...)
- Free tier: 100 experiments/month to prototype upsell tests
Drafting upsell subject lines
Upsell emails live or die on the subject line — the recipient already uses your product, so the angle has to feel relevant rather than promotional. In Continue, describe the upgrade and the audience, and ask the model to call generate_variants for a spread of angles you can test against each other.
Good upsell variants usually pull on a small set of distinct levers. Ask for variety so your subject line test isn't just rewording the same idea.
- Value framing: "You're close to your plan limit — here's more headroom"
- Feature unlock: "Turn on advanced reports with Pro"
- Usage-based proof: "You've sent 8,000 emails this month"
- Time-bound nudge: "Your upgrade discount ends Friday"
Running the subject line A/B test
With variants drafted, have the assistant call launch_test to send them as a managed experiment. marginal handles delivery and tracks opens and clicks, then reports within-test lift so you can see which upsell line moves the needle on your segment.
Because everything runs through the MCP tools, the loop stays inside Continue: launch a subject line A/B test, pull results, and let recommend_next suggest where to take the following round.
- launch_test — send your upsell variants as a tracked experiment
- get_results — read open/click rates and within-test lift
- recommend_next — get a data-backed suggestion for the next variant
An AI agent workflow for upsell email
Treating subject line testing as an agent task changes the cadence. Instead of manually exporting reports, you ask Continue for the latest numbers mid-conversation and iterate immediately — AI agent email marketing where the agent owns generation, sending, and analysis.
Over a few cycles you build a repeatable pattern: generate fresh upsell angles, ship a subject line test, review lift, then double down on the winning frame for your next campaign. The marginal docs at https://marginal.sh/docs/ cover each tool in detail when you want 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.
- 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