Using recommend_next to plan your next email test
The recommend_next email test tool is the part of marginal that turns a finished experiment into a smarter follow-up. Instead of guessing what to try next, your AI agent reads the lift data and proposes the next subject line A/B test automatically.
What recommend_next actually does
recommend_next is one of four tools exposed by the marginal email MCP server, alongside generate_variants, launch_test, and get_results. Once a test has collected enough opens and clicks, recommend_next analyzes the within-test lift and suggests how to design the follow-up experiment.
It is built for iteration. Rather than treating each subject line A/B test as a one-off, recommend_next looks at which variant won and why, then points your agent toward the angle, length, or tone worth testing next.
- Reads results from a completed or running experiment
- Surfaces the winning variant and its measured lift
- Proposes the next test direction based on observed performance
- Keeps the optimization loop running without manual analysis
Where it fits in the test loop
A typical cycle inside an AI agent email marketing workflow looks like a chain of tool calls. recommend_next sits at the end of that chain and feeds the beginning of the next one, so your campaigns keep improving over time.
- generate_variants — draft several subject line options
- launch_test — split traffic and start the managed send
- get_results — pull open and click metrics with lift
- recommend_next — decide what the next email test should be
Why agents handle this well
Deciding what to test next is the kind of judgment task LLM agents are good at when grounded in real numbers. Because recommend_next returns structured guidance, your agent in Cursor, Claude Desktop, Claude Code, Windsurf, or any MCP client can act on it directly — generating the next batch of variants and launching another subject line A/B test in the same conversation.
marginal is fully hosted, so there is no infrastructure to maintain. You connect to the endpoint at https://marginal.sh/mcp with a Bearer API key and the recommend_next tool is available immediately.
Getting started
Point your MCP client at https://marginal.sh/mcp, authenticate with your marg_live_ key, and run a first experiment. After get_results returns lift data, call recommend_next to plan the follow-up. The free tier includes 100 experiments per month, which is plenty for building a continuous testing habit.
See the full tool reference and example flows in the docs at https://marginal.sh/docs/.
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