Make.com Email Testing MCP
If you orchestrate marketing automation in Make.com but want an AI agent to actually decide and run your subject line experiments, the Make.com email testing MCP pattern with marginal closes that gap — a hosted email MCP server that exposes testing as callable tools.
Why pair Make.com with an email MCP server
Make.com is great at moving data between apps and triggering scenarios on a schedule or webhook. What it doesn't do natively is reason about which subject line will win, generate fresh variants, or interpret open and click results. That's where marginal comes in.
marginal is a hosted email MCP server built for AI agents. Instead of wiring up brittle modules to compute lift by hand, you let an agent call structured tools to generate variants, launch a test, and read back results — then feed the winner back into your Make.com flow.
- Make.com handles triggers, scheduling, and downstream actions
- marginal handles variant generation, A/B testing, and lift measurement
- The AI agent connects to both and makes the call on what to send next
The four MCP tools your agent will use
marginal exposes a focused set of tools over the MCP endpoint at https://marginal.sh/mcp. An AI agent embedded in your stack can chain them to run a complete subject line A/B test without a human writing the variants or crunching the stats.
- generate_variants — produce candidate subject lines from a brief
- launch_test — start a managed send and split traffic across variants
- get_results — pull open/click tracking and within-test lift metrics
- recommend_next — get the suggested winner or next action to take
A practical workflow for AI agent email marketing
A common setup: a Make.com scenario fires when a new campaign brief lands (a form submission, a CRM update, a scheduled run). That event reaches an AI agent that calls generate_variants and launch_test on marginal, then waits before calling get_results.
Once the subject line A/B test has enough signal, the agent calls recommend_next and pushes the winning subject back into Make.com — to log it in a sheet, notify a channel, or queue the broadcast send. The agent owns the testing logic; Make.com owns the plumbing.
- Agent authenticates to marginal with a Bearer API key (marg_live_...)
- Tracking covers opens and clicks so lift is measured within the test
- The free tier includes 100 experiments per month to validate the loop
Getting connected
marginal works with MCP-capable clients including Cursor, Claude Desktop, Claude Code, OpenAI Codex, Windsurf, Cline, Continue, and Zed. Point any of them at the marginal endpoint, add your API key, and the testing tools become available to your agent — which can then be triggered from or report back to your Make.com scenarios.
Because marginal is fully hosted, there's no self-hosted MCP server to deploy or maintain. See the docs at https://marginal.sh/docs/ for tool schemas and authentication details.
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