Using a VS Code MCP Email Server with marginal
If you live in VS Code and want your AI assistant to run real email experiments, marginal gives you a hosted email MCP server that plugs straight into your MCP-compatible extensions. No infrastructure to host — just an API key and a few tools your agent can call.
What a VS Code MCP email server gives you
VS Code has become a hub for AI coding assistants that speak the Model Context Protocol. When you point one of those assistants at marginal, the same chat panel you use to refactor code can now design and ship subject line experiments. The agent calls remote tools over HTTPS; marginal handles the sending, tracking, and analysis.
- generate_variants — draft multiple subject lines for one campaign
- launch_test — start a subject line A/B test with a managed send
- get_results — pull open and click data plus within-test lift
- recommend_next — get a data-backed suggestion for your next test
Which VS Code clients work
marginal is a standard MCP server, so any MCP-capable client that runs inside or alongside VS Code can reach it. That includes Cline and Continue as VS Code extensions, plus editor-native options like Cursor, Windsurf, and Zed. Claude Code, Claude Desktop, and OpenAI Codex round out the supported clients if you switch contexts.
Because the server is hosted at https://marginal.sh/mcp, configuration is the same across clients: register the remote endpoint and supply a Bearer API key. There is no self-hosted MCP server to maintain and no background process to babysit.
- Cline and Continue — MCP support directly in VS Code
- Cursor, Windsurf, Zed — editor-native MCP clients
- Claude Code and Claude Desktop — terminal and desktop agents
- OpenAI Codex — agentic workflows outside the editor
Running your first subject line A/B test
Once your VS Code MCP email server connection is live, a typical flow is conversational. Ask your agent to generate a handful of subject lines for a campaign, launch the test, then check back later for results. The agent chains the tools for you, so you never leave the editor.
Every test returns within-test lift metrics, so you can see which variant actually moved opens and clicks rather than guessing. From there, recommend_next helps you turn one result into a learning loop instead of a one-off.
- "Generate five subject lines for our onboarding email"
- "Launch an A/B test across those variants"
- "Show me open and click lift once data is in"
- "What should we test next?"
Getting set up
Grab a marg_live_ API key, add the https://marginal.sh/mcp endpoint to your VS Code MCP client, and start experimenting. The free tier covers 100 experiments per month, which is plenty to validate the AI agent email marketing workflow before scaling up.
Full configuration steps for each client live in the docs at https://marginal.sh/docs/, and the MCP overview at https://marginal.sh/mcp lists the registry name sh.marginal/mcp for clients that resolve servers by identifier.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer key in the marg_live_ format
- Registry: sh.marginal/mcp
- Docs: 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