Using the Continue dev email MCP with marginal
If you build inside Continue, you can hand subject line testing and managed email sends to your agent. The marginal email MCP server connects to Continue over a single hosted endpoint, so your assistant can generate variants, launch tests, and read lift metrics without leaving your editor.
Why pair Continue with an email MCP server
Continue is an open-source AI code assistant that supports the Model Context Protocol, which means it can call external tools that follow the MCP spec. marginal exposes email marketing as exactly that kind of tool surface — a hosted email MCP server at https://marginal.sh/mcp.
Instead of stitching together send APIs and analytics dashboards by hand, you give Continue a set of structured tools. The agent decides when to draft new subject lines, when to launch a test, and when enough data has arrived to call a winner.
- generate_variants — produce candidate subject lines to test
- launch_test — start a managed subject line A/B test
- get_results — pull open, click, and within-test lift metrics
- recommend_next — get a data-backed suggestion for the next step
What the Continue dev email MCP setup looks like
Because marginal is fully hosted, there is nothing to deploy. You point Continue at the marginal endpoint and authenticate with a Bearer API key in the form marg_live_…. No self-hosted MCP server, no separate send infrastructure to maintain.
Once connected, your Continue session can treat email experimentation as part of its normal toolset. Ask it to draft variants for a launch announcement, and it can call generate_variants, then launch_test, and check back later with get_results.
- Registry: sh.marginal/mcp
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Free tier: 100 experiments per month
- Full setup steps live in the docs at https://marginal.sh/docs/
A subject line A/B test, agent-driven
A practical workflow for AI agent email marketing in Continue starts with intent: you describe the email and the audience, and the agent does the mechanical work. The marginal tools keep each step deterministic and observable.
A typical loop runs like this:
- Generate three to five subject line variants from your brief
- Launch a subject line A/B test across your list with managed sends
- Track opens and clicks as results accumulate
- Read within-test lift metrics to compare variants fairly
- Let recommend_next surface the winner or suggest a follow-up test
Tips for reliable results in Continue
Treat the agent as a fast operator, not an oracle. Give it clear constraints — tone, length, audience — so the variants it generates are worth testing. Then let the data decide.
Lean on lift metrics rather than raw opens when comparing variants, since within-test measurement controls for timing and audience overlap. The marginal email MCP server reports these directly, so Continue can reason over them without extra glue code.
- Keep variant sets small so each gets enough traffic
- Use get_results before recommend_next so decisions are grounded in data
- Mind the 100-experiment monthly free tier when scripting many runs
- Store your marg_live_ key securely in Continue's MCP config
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