Reading the get_results Email Experiment Output
Once you've launched a subject line A/B test through marginal, the get_results tool is how your AI agent reads the scoreboard. It returns within-test lift metrics, open rates, and click data so an agent can decide what to send next without you babysitting the dashboard.
What get_results returns
get_results is one of four MCP tools exposed by marginal, alongside generate_variants, launch_test, and recommend_next. After a test has collected enough sends, calling get_results on the experiment ID gives your agent a structured snapshot of how each variant is performing.
- Open and click counts per variant
- Open and click rates normalized by audience size
- Within-test lift — how each variant compares to the control
- Experiment status so an agent knows when results are stable
- Variant identifiers that map back to the subject lines you tested
How an AI agent uses it in a loop
The get_results email experiment call is the measurement step in a tight feedback cycle. An AI agent generates variants, launches the test, polls results, and then asks recommend_next for guidance — repeating until a clear winner emerges.
Because marginal is a hosted email MCP server, your agent talks to a single endpoint at https://marginal.sh/mcp with a Bearer API key. There's nothing to deploy and no self-hosted infrastructure to keep running between experiments.
- generate_variants — draft candidate subject lines
- launch_test — start a managed subject line A/B test
- get_results — pull lift, opens, and clicks
- recommend_next — decide the winning variant or iterate
Tips for reliable readings
Lift numbers are only meaningful once a test has accumulated enough opens and clicks. Have your agent treat early get_results responses as provisional and wait for the status to settle before declaring a winner.
- Poll get_results on an interval rather than immediately after launch
- Compare within-test lift, not raw counts, across variants
- Let recommend_next translate the numbers into an action
- Use the free tier's 100 experiments/month to validate your flow
- See the docs at https://marginal.sh/docs/ for the full response shape
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