An Email MCP Server for Agencies
If your agency runs email campaigns for a roster of clients, marginal gives your AI tools a single hosted endpoint to generate variants, launch tests, and report lift — without standing up infrastructure per account.
Why agencies need an email MCP server
Agencies juggle dozens of campaigns, deadlines, and brand voices at once. An email MCP server for agencies lets your AI assistants do the repetitive testing work — drafting subject line variants, launching experiments, and pulling results — directly from the editor or chat where your team already works.
marginal is hosted, so there's no self-hosted server to maintain and no per-client deployment to babysit. You connect once through https://marginal.sh/mcp and your agents call the same four tools across every account you manage.
- One endpoint instead of bespoke scripts per client
- Managed sends with open and click tracking included
- Within-test lift metrics so wins are defensible to clients
- Free tier of 100 experiments/month to evaluate workflows
Running a subject line A/B test from an AI agent
The core agency loop is fast with marginal. An AI agent calls generate_variants to produce subject line candidates, launch_test to send the split, get_results to read open and click performance, and recommend_next to decide the follow-up. Each step is a tool call your assistant can chain on its own.
Because every subject line A/B test reports within-test lift, you can show a client exactly how much a winning line outperformed the control instead of guessing from raw open rates.
- generate_variants — draft on-brand subject line options
- launch_test — send the A/B split with managed delivery
- get_results — pull open/click metrics and lift
- recommend_next — get the suggested next experiment
Fits the tools your team already uses
AI agent email marketing only sticks when it lives inside daily workflows. marginal works as an MCP server in the clients agency teams reach for, so a copywriter in Claude and an engineer in Cursor can both trigger tests against the same account.
Authentication is a single Bearer API key (marg_live_...). Add the endpoint to your MCP client config, drop in the key, and your agents can start running experiments.
- Cursor, Claude Desktop, Claude Code, and Windsurf
- OpenAI Codex, Cline, Continue, and Zed
- Registry: sh.marginal/mcp · Endpoint: https://marginal.sh/mcp
- Docs and setup details at https://marginal.sh/docs/
Scaling tests across a client roster
As an agency grows, the bottleneck is rarely ideas — it's the time to execute and measure them. With marginal as your email MCP server, agents handle the mechanical parts of every campaign so your team spends its hours on strategy and client relationships.
Start small on the free tier to prove the loop on one or two accounts, then expand the same setup across the rest of your roster without changing your tooling.
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