Product Launch Subject Lines in Claude Code
A product launch lives or dies by the open rate of its announcement email. This guide shows how to draft, test, and pick winning product launch subject lines in Claude Code by connecting it to marginal — a hosted email MCP server that handles variant generation, managed sends, and within-test lift metrics.
Why test launch subject lines from Claude Code
When you're shipping a product, the announcement email is often the single highest-leverage message you'll send all quarter. Claude Code already sits in your workflow alongside release notes, changelogs, and copy drafts — so it's a natural place to spin up a subject line test without context-switching into a separate marketing tool.
With marginal connected as an MCP server, Claude Code can call email-marketing tools directly. You describe the launch, the agent generates subject line candidates, runs the A/B test, and reports back which variant won — all in the same session where you wrote the feature.
- Draft launch copy and test subject lines in one conversation
- Keep variants tied to the actual feature you just shipped
- Let the AI agent handle the email marketing mechanics end to end
The marginal tools behind a subject line A/B test
marginal exposes four MCP tools that map cleanly onto a launch test. Claude Code orchestrates them in sequence, so a single prompt like "test three subject lines for our v2 launch" can carry through generation, send, and analysis.
- generate_variants — produce product launch subject line candidates from your prompt
- launch_test — start the subject line A/B test against your audience with managed sends
- get_results — pull open and click tracking plus within-test lift metrics
- recommend_next — get a data-backed suggestion for the winning variant or a follow-up
A practical launch workflow
A typical product launch subject line test in Claude Code follows a short loop. Start by giving the agent your launch angle — the feature, the benefit, and the tone you want. Ask it to call generate_variants for several distinct hooks (curiosity, benefit-led, urgency). Review the candidates inline, trim the list, then have it call launch_test.
Once sends are out, ask Claude Code to call get_results. You'll see opens, clicks, and the lift each variant earned within the test, so you can confirm a winner with evidence rather than a guess. From there, recommend_next can point you to the variant to scale or a refinement worth a second round.
Connecting Claude Code to marginal
marginal is fully hosted — there's no self-hosted server to run. Point Claude Code at the endpoint at https://marginal.sh/mcp and authenticate with your Bearer API key (the marg_live_... token from your dashboard). The free tier covers 100 experiments per month, which is plenty for iterating on a launch announcement.
Once it's wired up, AI agent email marketing becomes part of your normal coding session. See the setup details and tool reference at https://marginal.sh/docs/ to get the connection running.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Registry: sh.marginal/mcp
- Docs: 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