A Developer Friendly Email Platform Built for AI Agents
If you've ever wanted to run a subject line test from your editor instead of clicking through a marketing dashboard, that's the gap marginal fills. It's a developer friendly email platform delivered as a hosted MCP server, so the same agent that writes your copy can also launch the test.
What makes a platform developer friendly?
Most email tools are built for marketers first and developers second. That usually means a clickable UI, exported reports, and an API bolted on afterward. marginal flips the order: the primary interface is an email MCP server that your tools talk to directly, with structured tools instead of brittle dashboard scraping.
Authentication is a single Bearer API key (marg_live_...), the endpoint is one URL, and the responses are machine-readable JSON your agent can act on. No SDK lock-in, no self-hosting to maintain — the server is hosted at marginal.sh.
- Single endpoint: https://marginal.sh/mcp
- Bearer key auth — no OAuth dance to script around
- Hosted and managed — nothing to deploy or patch
- Free tier of 100 experiments/month to prototype against
Subject line A/B testing without leaving your editor
The core workflow is a subject line A/B test. Your agent generates candidate variants, launches a managed send, and reads back open and click metrics with within-test lift so you can see which line actually won — not just which one looked good in isolation.
Because every step is an MCP tool, you can wire this into existing automations: a PR triggers copy generation, a scheduled job launches the test, and a follow-up reads results to pick a winner.
- generate_variants — draft subject line options
- launch_test — start a managed A/B send
- get_results — open/click tracking with lift metrics
- recommend_next — get a data-backed suggested next step
AI agent email marketing, end to end
marginal is designed for AI agent email marketing where the loop runs without a human babysitting a dashboard. The agent proposes, tests, measures, and iterates — and the lift metrics keep those iterations grounded in real recipient behavior rather than guesswork.
It works with the MCP clients developers already use, so adding email testing to an agent is a config change, not a migration.
- Cursor, Claude Code, Claude Desktop, and Windsurf
- OpenAI Codex, Cline, Continue, and Zed
- Same tools across every client via the MCP standard
Getting started
Point your MCP client at the marginal endpoint, add your API key, and the four tools appear in your agent's toolset. From there, ask it to draft and launch your first subject line A/B test.
Full setup details and per-client instructions live in the docs at https://marginal.sh/docs/, and the MCP connection reference is at https://marginal.sh/mcp.
- Registry: sh.marginal/mcp
- Docs: https://marginal.sh/docs/
- Start on the free tier with 100 experiments per 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