Browse Abandonment Subject Lines in Continue
Browse abandonment emails recover shoppers who viewed products but never added to cart. With Continue connected to marginal, you can draft and A/B test those subject lines from the same editor where you build the rest of your stack.
Why pair Continue with browse abandonment campaigns
Browse abandonment is a high-frequency trigger: most visitors browse far more than they buy, so the subject line does the heavy lifting in deciding whether a recovery email gets opened at all. Continue's agent workflow lets you iterate on those subject lines conversationally instead of hopping between tools.
marginal exposes its email MCP server to Continue so the agent can generate subject line variants, launch a test, and report results. You stay in your editor and the agent handles the send mechanics.
- Browse abandons happen earlier in the funnel than cart abandons, so curiosity-driven subject lines often outperform urgency framing.
- Continue's chat-first interface makes rapid subject line iteration feel natural.
- Open and click tracking come back into the same conversation for fast review.
Running a subject line A/B test from Continue
Once marginal is registered as an MCP server, ask Continue to draft browse abandonment subject lines around a product category or recently viewed item. The agent calls generate_variants, then launch_test to split your audience across the options.
A clean subject line A/B test isolates one variable at a time. For browse abandonment, that usually means comparing tone — product reminder versus social proof versus a soft incentive — rather than changing several elements at once.
- generate_variants — produce a set of browse abandonment subject lines to compare.
- launch_test — split traffic and start the managed send.
- get_results — pull opens, clicks, and within-test lift metrics.
- recommend_next — let the agent suggest the next variant to try.
Reading lift and deciding the winner
After the send, ask Continue for results and marginal returns open and click data with within-test lift so you can see which subject line moved browsers toward the catalog. Because the comparison is internal to the test, you avoid drawing conclusions from noisy week-over-week swings.
Use recommend_next to keep the loop going — feed the winning angle into a fresh subject line test for the next browse abandonment flight and compound your learnings.
- Look at click-through, not just opens, since browse intent is weaker than cart intent.
- Promote the winner, then test a new dimension against it.
- The free tier covers 100 experiments per month for steady iteration.
Connecting marginal to Continue
marginal is a hosted MCP server, so there's nothing to self-host. Point Continue at the endpoint https://marginal.sh/mcp and authenticate with a Bearer API key (marg_live_...). From there the AI agent email marketing workflow runs entirely through MCP tool calls.
See the docs at https://marginal.sh/docs/ for the full tool reference, or grab the endpoint details at https://marginal.sh/mcp to get Continue talking to marginal.
- Registry: sh.marginal/mcp
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Tools: generate_variants, launch_test, get_results, recommend_next
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