A startup email A/B testing tool your AI agent can drive
If you're shipping fast and shouldn't be hand-tuning email campaigns, marginal gives you a startup email A/B testing tool that runs entirely through your AI coding agent — generate variants, launch a subject line A/B test, and read lift metrics without leaving your editor.
Why startups need a leaner A/B testing workflow
Early-stage teams rarely have a dedicated email marketer. The person writing onboarding emails is often the same engineer building the product, and the overhead of a traditional ESB campaign builder slows everything down.
marginal removes that friction by exposing email experimentation as MCP tools. Instead of clicking through a dashboard, you describe the test you want and your agent calls the underlying tools to make it happen.
- No campaign-builder UI to learn — drive everything from your agent
- Free tier of 100 experiments/month, enough for most pre-launch and seed-stage sends
- Open and click tracking baked in, so you get real signal on what works
How a subject line A/B test works in marginal
A subject line A/B test in marginal is a four-step loop. You generate candidate variants, launch the test against your list, fetch results once data comes in, and ask for a recommendation on what to send next.
Each step maps to a single MCP tool, so an AI agent email marketing workflow can chain them together autonomously or pause for your approval at any stage.
- generate_variants — produce subject line options to test
- launch_test — start a managed send and split the audience
- get_results — pull open/click rates and within-test lift metrics
- recommend_next — get a data-backed call on the winning variant
Connecting the email MCP server to your agent
marginal is a hosted email MCP server, so there's nothing to deploy or self-host. You point your MCP client at https://marginal.sh/mcp and authenticate with a Bearer API key in the marg_live_ format.
The server is registered as sh.marginal/mcp and works with the MCP clients startups already use — Cursor, Claude Code, Claude Desktop, OpenAI Codex, Windsurf, Cline, Continue, and Zed.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Docs and setup guides: https://marginal.sh/docs/
Putting it into practice
Treat marginal like any other tool in your agent's toolbox. Ask it to draft three subject lines for a launch announcement, launch the A/B test, and report back the winner once enough recipients have engaged.
Because results come with within-test lift metrics, you can make decisions on actual performance rather than gut feel — which is exactly what a startup email A/B testing tool should make easy.
- Iterate on subject lines without leaving your IDE
- Let the agent recommend the next send based on real data
- Scale up experiments as your list and send volume grow
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