Enterprise Email Experimentation MCP
An enterprise email experimentation MCP lets your AI agents design, launch, and evaluate subject line tests without bouncing between dashboards. marginal is a hosted email MCP server that exposes experimentation as callable tools, so the same agent that drafts your campaign can also measure which version wins.
What an experimentation MCP gives enterprise teams
At scale, the bottleneck isn't sending email — it's keeping experimentation consistent across teams, agents, and campaigns. An enterprise email experimentation MCP standardizes how variants get created and how results get read, so an AI agent email marketing workflow behaves the same way every time it runs.
marginal handles the mechanics: managed sends, open and click tracking, and within-test lift metrics. Your agents focus on strategy and copy while the MCP server runs the controlled test underneath.
- Consistent test methodology across every agent and campaign
- Bearer API key auth (marg_live_...) for clean access control
- Managed sends with built-in open and click tracking
- Within-test lift metrics instead of raw, hard-to-compare numbers
The four tools behind a subject line A/B test
marginal exposes a focused set of MCP tools. An agent calls them in sequence to take a subject line A/B test from idea to decision, and each tool returns structured output the next step can act on.
- generate_variants — draft multiple subject line candidates from a prompt
- launch_test — start a managed A/B test across your audience split
- get_results — pull opens, clicks, and lift figures as data lands
- recommend_next — get a data-backed suggestion for the follow-up send
Built into the tools your engineers already use
Because marginal is a hosted email MCP server reachable at https://marginal.sh/mcp, you connect once and call it from any MCP-capable client. There's nothing to self-host and no infrastructure to maintain — point your agent at the endpoint and the experimentation tools appear.
That makes it easy to embed experimentation into developer and growth workflows without a separate marketing console.
- Works with Cursor, Claude Desktop, Claude Code, and OpenAI Codex
- Also supported in Windsurf, Cline, Continue, and Zed
- Registry identifier: sh.marginal/mcp
- Reference docs at https://marginal.sh/docs/
Start experimenting at low risk
The free tier includes 100 experiments per month, which is enough to validate that an MCP-driven experimentation loop fits how your teams work before scaling it across campaigns. Run a few real subject line A/B tests, confirm the lift metrics line up with your expectations, and let your AI agent email marketing process take it from there.
- 100 experiments per month on the free tier
- Endpoint: https://marginal.sh/mcp
- No self-hosted server to operate — marginal is fully hosted
- Same tools power both quick tests and ongoing enterprise programs
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