A 250k Recipients Email Test Platform Built for AI Agents
When your list reaches the quarter-million mark, guessing at subject lines stops being acceptable. marginal is a hosted email MCP server that lets your AI agent generate variants, launch tests, and read lift metrics — all from inside your editor or coding agent.
Why list size changes how you test
At 250k recipients, a single percentage point of open-rate lift moves thousands of real opens. That scale rewards disciplined experimentation: split your audience cleanly, hold everything constant except the subject line, and measure the difference with statistical honesty.
marginal handles the mechanics so you can focus on the hypothesis. A subject line A/B test splits a representative slice of your 250k recipients, sends each variant, and reports open and click performance with within-test lift — no spreadsheet wrangling required.
- Clean audience splits sized for large lists
- Open and click tracking on every variant
- Within-test lift metrics so winners are obvious
- Managed sends — no separate delivery setup to babysit
How the email MCP server fits your workflow
marginal exposes its testing tools over the Model Context Protocol at https://marginal.sh/mcp. That means your AI agent calls the same four tools you would: generate_variants to draft subject lines, launch_test to send the split, get_results to pull metrics, and recommend_next to suggest the follow-up experiment.
Because marginal is fully hosted, there is no self-managed infrastructure to maintain at scale. You authenticate with a Bearer API key (marg_live_...) and your agent does the rest — even when the test fans out across a 250k recipients email test platform workload.
- generate_variants — produce candidate subject lines
- launch_test — run the A/B split and send
- get_results — fetch opens, clicks, and lift
- recommend_next — plan the next iteration
AI agent email marketing at scale
marginal connects to the tools developers already use: Cursor, Claude Desktop, Claude Code, OpenAI Codex, Windsurf, Cline, Continue, and Zed. Your agent reads results and proposes the next subject line variant without you leaving the editor, turning AI agent email marketing into a tight, repeatable loop.
This closes the gap between idea and evidence. Instead of one big quarterly send, you can run frequent subject line A/B tests, learn from each one, and let recommend_next compound those learnings across your large list.
Getting started
The free tier includes 100 experiments per month — enough to validate subject line testing against a meaningful portion of a 250k-recipient program before scaling up. Point your MCP client at https://marginal.sh/mcp, add your API key, and ask your agent to run the first test.
Full tool references, payload examples, and setup steps for each supported client live in the docs at https://marginal.sh/docs/.
- Registry: sh.marginal/mcp
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Free tier: 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