Email Analytics for AI Workflows
If your AI agents are already drafting and sending email, the missing piece is measurement. marginal exposes email analytics for AI workflows directly through an MCP server, so an agent can launch a test, read open and click data, and decide what to do next — without you exporting CSVs or stitching together dashboards.
Why analytics belongs inside the agent loop
Most email tooling treats analytics as a human dashboard problem. That breaks down when an AI agent is the one making decisions, because the agent needs structured numbers it can act on, not a chart it has to screenshot.
marginal closes that gap by returning metrics as tool output the model can reason over. The same loop that wrote the copy can also ask how it performed and adjust.
- Open and click tracking returned as machine-readable results
- Within-test lift metrics so the agent can compare variants directly
- No manual export step between sending and analyzing
- Works inside the agent's existing context window
The MCP tools that power the analytics loop
marginal is an email MCP server exposing four focused tools. Together they cover the full cycle of AI agent email marketing: write, test, measure, iterate.
- generate_variants — produce subject line and copy options to test
- launch_test — start a subject line A/B test with a managed send
- get_results — pull open/click metrics and within-test lift
- recommend_next — turn the analytics into a concrete next action
A typical subject line A/B test, end to end
Because the analytics live next to the sending tools, a subject line A/B test runs as one continuous conversation. The agent generates variants, launches the test, waits for engagement to accumulate, then reads results and acts.
This is what makes email analytics for AI workflows practical instead of theoretical: the data feeds back into the same agent that produced the experiment, closing the loop automatically.
- Agent calls generate_variants for several subject lines
- launch_test sends them as a managed A/B test
- get_results reports opens, clicks, and lift per variant
- recommend_next suggests the winner or a follow-up test
Getting connected
marginal is hosted, so there's nothing to deploy or maintain. Point any MCP-capable client — Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Codex, or Zed — at https://marginal.sh/mcp and authenticate with a Bearer API key (marg_live_...).
The free tier covers 100 experiments per month, which is enough to validate an analytics-driven workflow before scaling. Full tool references live in the docs at https://marginal.sh/docs/.
- Endpoint: https://marginal.sh/mcp
- Registry: sh.marginal/mcp
- Auth: Bearer API key
- 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