Cross-Sell Subject Line Tests in Zed
Zed's AI agent can do more than write code. Wire it to marginal — a hosted email MCP server — and you can draft, test, and measure cross-sell subject lines straight from your editor.
Why run cross-sell subject line tests from Zed
Cross-sell campaigns live or die on the subject line. The exact same product recommendation lands very differently depending on whether you lead with a benefit, a bundle discount, or a curiosity hook. Rather than guessing, you can let Zed's agent propose several angles and put them head-to-head.
Because marginal exposes its tooling over MCP, the work happens where you already are. No browser tab switching, no copy-pasting subject lines into a separate campaign dashboard — you describe the cross-sell, and the agent handles variant generation and the subject line A/B test.
- Keep your campaign workflow inside the Zed agent loop
- Generate cross-sell angles and test them in one conversation
- Read open/click lift without leaving the editor
Connecting marginal to Zed
Add marginal as an MCP server in Zed using the hosted endpoint at https://marginal.sh/mcp. Authentication is a Bearer API key in the form marg_live_..., which you generate from your account. marginal is fully hosted, so there's nothing to deploy or self-host — Zed just connects to the registry entry sh.marginal/mcp.
Once connected, the agent can call four tools: generate_variants, launch_test, get_results, and recommend_next. Those cover the full loop from idea to decision for any cross-sell subject line test.
- Endpoint: https://marginal.sh/mcp
- Registry: sh.marginal/mcp
- Auth: Bearer API key (marg_live_...)
- Free tier: 100 experiments/month
- Docs: https://marginal.sh/docs/
A cross-sell test workflow in practice
Say a customer just bought a standing desk and you want to cross-sell a monitor arm. In Zed, prompt the agent with that context and ask for several subject line angles. It calls generate_variants to produce options — a value framing, a compatibility framing, a scarcity framing — then launch_test sends them as a managed A/B test.
As opens and clicks accumulate, get_results returns within-test lift metrics so you can see which framing actually moved the needle. When you're ready, recommend_next suggests the follow-on move based on what won.
- Describe the cross-sell offer and audience to the Zed agent
- generate_variants drafts distinct subject line angles
- launch_test runs the managed subject line A/B test
- get_results shows open/click lift within the test
- recommend_next points to the strongest next step
Tips for AI agent email marketing
Cross-sell subject lines reward specificity. Give the agent the original purchase, the recommended product, and the relationship between them — that produces sharper variants than a generic prompt. Treat each test as a small experiment and let the data, not your instincts, pick the winner.
Keep your tests scoped: a focused two-to-four variant subject line test resolves faster and gives cleaner lift signals than testing a dozen at once. With AI agent email marketing in Zed, the cost of iterating is low, so run more tests rather than fewer.
- Feed the agent both the past purchase and the cross-sell item
- Limit each test to a handful of clearly different angles
- Use within-test lift, not raw opens, to declare a winner
- Iterate often — the free tier covers 100 experiments/month
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