Subject Line A/B Testing in Continue
Continue brings agentic coding and chat right into your editor. Connect it to marginal — a hosted email marketing MCP server — and you can design, launch, and analyze a subject line A/B test without leaving your workspace.
Why run subject line A/B tests from Continue
Continue's agent workflow lets you chain reasoning and tool calls together, which maps cleanly to how a subject line A/B test actually works: draft candidates, send them to segments, then compare opens and clicks. With marginal wired in as an MCP server, Continue can do all of that through plain instructions.
Because marginal is hosted, there's nothing to deploy and no infrastructure to babysit. You point Continue at the endpoint, authenticate with a Bearer key, and the four tools become available to the agent.
- Keep the entire subject line A/B test inside your editor — no context switching to a separate dashboard.
- Let the Continue agent workflow handle the repetitive steps of variant generation and result polling.
- Iterate fast: tweak a prompt, regenerate variants, relaunch.
Connecting marginal to Continue
Add marginal as an MCP server in Continue's configuration. The endpoint is https://marginal.sh/mcp and authentication uses a Bearer API key that starts with marg_live_. Once registered, Continue can call marginal's tools during any chat or agent session.
- Endpoint: https://marginal.sh/mcp
- Registry name: sh.marginal/mcp
- Auth: Bearer API key (marg_live_...)
- Tools exposed: generate_variants, launch_test, get_results, recommend_next
A working subject line A/B test workflow
Once connected, describe what you want and let the agent drive the tools in order. A typical AI agent email marketing loop in Continue looks like this:
- Call generate_variants to produce several subject line candidates for your campaign.
- Use launch_test to send the variants as a managed split with open and click tracking enabled.
- Poll get_results to read within-test lift metrics as data accumulates.
- Ask recommend_next to decide the winner or refine for the next round.
Tips for cleaner results in Continue
A subject line A/B test is only as good as its inputs. Give Continue enough campaign context — audience, tone, and the offer — so generate_variants produces distinct options rather than near-duplicates. Then let the lift metrics, not gut feeling, pick the winner.
The free tier covers 100 experiments per month, which is plenty for regular testing while you build out your Continue agent workflow. See the docs at https://marginal.sh/docs/ for tool schemas and examples.
- Wait for a meaningful sample before reading get_results.
- Test one variable at a time so the subject line is the only difference.
- Feed recommend_next back into your next generate_variants prompt to compound learnings.
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