Streamable HTTP MCP Email: Run Subject Line Tests from Any Agent
If you're searching for a streamable HTTP MCP email setup, you want an agent that can generate variants, launch a send, and read results without juggling SMTP libraries or vendor SDKs. marginal is a hosted email MCP server that speaks the streamable HTTP transport, so any Model Context Protocol client can drive your subject line A/B test from a single URL.
What streamable HTTP MCP means for email
The Model Context Protocol supports a streamable HTTP transport, which lets an MCP client talk to a remote server over plain HTTP requests instead of spawning a local process. For email work, that's the difference between maintaining a self-hosted daemon and simply pointing your agent at a hosted endpoint.
marginal exposes its tools at https://marginal.sh/mcp using exactly this transport. There is no local server to install or keep running — your agent connects, authenticates with a Bearer API key, and starts orchestrating campaigns.
- Endpoint: https://marginal.sh/mcp
- Transport: streamable HTTP (remote MCP)
- Auth: Bearer API key in the form marg_live_...
- No self-hosted process — marginal is fully hosted
The email MCP tools your agent calls
Once connected, an AI agent has four focused tools to plan and execute a subject line A/B test. Each maps to a concrete step in the experiment loop, so the model can reason about what to do next instead of stitching together raw API calls.
- generate_variants — draft competing subject lines for a campaign
- launch_test — send the variants and split your audience
- get_results — pull open and click tracking plus within-test lift
- recommend_next — get a data-backed suggestion for the follow-up send
A typical AI agent email marketing loop
Because the streamable HTTP MCP email server returns structured results, your agent can run a full test cycle in one conversation. Ask it to generate three subject lines, launch the test to a sample, wait for engagement data, then act on the lift metrics.
This makes AI agent email marketing practical rather than experimental: the agent measures real opens and clicks, identifies the winning variant within the test, and uses recommend_next to propose the next iteration.
- Generate variants for a launch announcement
- Launch the A/B test with managed sends and tracking
- Read within-test lift to find the winner
- Iterate with recommend_next on a fresh segment
Getting connected
marginal works with any MCP client that supports remote servers, including Cursor, Claude Desktop, Claude Code, OpenAI Codex, Windsurf, Cline, Continue, and Zed. Point your client at the endpoint, add your Bearer key, and the four email tools appear in the agent's toolset.
The free tier covers 100 experiments per month, which is enough to validate a real subject line A/B test workflow before you scale. Full setup details and the tool reference live in the docs at 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