← All guides marginal.sh

A subject line split test tool built for AI agents

Updated June 2026 · Hosted email MCP server

If you're searching for a subject line split test tool that your AI agent can actually drive, marginal exposes split testing as an MCP server. Your agent generates variants, launches the test, and reads results without you wiring up a dashboard.

What makes marginal a split test tool, not just an API

Most subject line testing lives behind a UI you click through manually. marginal flips that: it's a hosted email MCP server, so any MCP-capable client can call a small set of tools to run a full subject line A/B test end to end.

That means the loop — propose variants, send to segments, measure opens and clicks, pick a winner — happens in the same place your agent already works.

How a subject line A/B test runs

A typical split test starts with two or more subject line variants and a segment of recipients. marginal handles the managed send, randomized assignment, and tracking so you don't assemble the plumbing yourself.

As opens and clicks come in, get_results reports lift between variants within the same test, giving you an apples-to-apples comparison rather than guessing across separate campaigns.

Wiring it into AI agent email marketing

Because marginal speaks MCP, the split test tool plugs into the agents and editors you already use — Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, and OpenAI Codex.

Connect with a Bearer API key (marg_live_...) and point your client at https://marginal.sh/mcp. From there your agent can run subject line A/B tests as part of a broader campaign workflow, with no self-hosting to manage.

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.

Documentation · Sign up · FAQ