Testing shipping update subject lines in OpenAI Codex
Shipping update emails get opened — or ignored — based largely on the subject line. With marginal connected to OpenAI Codex over MCP, you can draft variants, launch a subject line test, and read the lift metrics straight from your coding workflow.
Why shipping updates need their own subject line test
Transactional and post-purchase shipping update emails sit in a tricky spot: customers expect them, but generic subjects like "Your order has shipped" leave engagement on the table. A subject line A/B test tells you which framing — order number, ETA, or carrier name — actually drives opens.
Because OpenAI Codex already understands the context of the order system you're building, it's a natural place to spin up these experiments alongside the code that triggers the send.
- Compare ETA-forward subjects ("Arriving Thursday") vs. status-forward ones ("Your package is on the way")
- Test whether including the order number lifts opens for repeat buyers
- Measure within-test lift so you ship the winner with confidence
Connecting OpenAI Codex to the marginal email MCP server
marginal is a hosted email MCP server, so there's nothing to self-host. Point OpenAI Codex at the endpoint, authenticate with a Bearer API key, and the four MCP tools become available to the agent.
Once connected, you can describe a shipping update scenario in plain language and let Codex call the tools for you.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key in the marg_live_... format
- Registry id: sh.marginal/mcp
- Tools exposed: generate_variants, launch_test, get_results, recommend_next
Running a shipping update subject line test end to end
The workflow inside OpenAI Codex mirrors how you'd reason about any experiment in code. Ask the agent to generate a handful of candidate subject lines for the shipping update, launch the test against your audience, and then poll for results once data accumulates.
This makes AI agent email marketing concrete: the agent does the busywork while you stay focused on the product, then surfaces a clear recommendation when the numbers settle.
- generate_variants — produce several shipping update subject line candidates
- launch_test — start the subject line A/B test and begin the managed send
- get_results — pull open/click tracking and within-test lift
- recommend_next — let marginal suggest the winning subject and follow-up step
Tips for better shipping update results
Keep each test focused on one variable at a time so the lift is interpretable. Shipping update audiences are reliable openers, which means you'll often reach significance faster than with promotional sends — a good reason to iterate frequently.
- Start on the free tier with up to 100 experiments per month
- Limit variants to 3–4 so each gets enough volume to read
- Re-test seasonally — delivery expectations shift around peak periods
- See https://marginal.sh/docs/ for tool schemas and examples
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