An Iterable Alternative for Agent-Driven Email Tests
If you want your AI agent to draft, launch, and measure subject line experiments without clicking through a campaign UI, marginal is a focused Iterable alternative for agent email tests — a hosted email MCP server that exposes testing as callable tools.
Why teams look past Iterable for agent workflows
Iterable is a full marketing automation suite, which is great if you live inside a visual workflow builder. But when your experiments are driven by an AI agent in Cursor or Claude Code, a heavyweight campaign platform becomes a bottleneck: the agent can't easily reason over a UI, and you end up scripting against broad REST endpoints.
marginal takes the opposite approach. It does one job — subject line A/B testing — and exposes it through the Model Context Protocol so an agent can run the entire loop programmatically.
- Purpose-built for subject line A/B test cycles, not full lifecycle orchestration
- Tools the agent calls directly: generate_variants, launch_test, get_results, recommend_next
- Hosted endpoint at https://marginal.sh/mcp — no self-hosted server to operate
- Free tier of 100 experiments per month to evaluate the workflow
How AI agent email marketing works with marginal
In an AI agent email marketing setup, the agent owns the experiment loop. It asks marginal to generate variants from a prompt or brief, launches a managed send to your audience split, then polls results and asks what to do next.
Because each step is a discrete MCP tool, the agent can chain them or pause for your approval. You stay in the loop without leaving your editor.
- generate_variants — produce candidate subject lines to test
- launch_test — start a managed send with the chosen split
- get_results — pull open/click tracking and within-test lift
- recommend_next — get a data-backed suggestion for the follow-up test
What you get on every subject line A/B test
marginal handles the send and the measurement so your agent only deals with decisions. Each subject line A/B test reports open and click tracking plus within-test lift metrics, so the winner is statistically grounded rather than guessed.
That makes iterating fast: run a test, read the lift, let the agent propose the next variant set, repeat.
- Managed sends — marginal delivers the email split for you
- Open and click tracking captured per variant
- Within-test lift metrics to compare performance fairly
- Bearer API key auth (marg_live_...) keeps agent calls scoped
Connecting your client to the email MCP server
marginal works as an email MCP server with any MCP-capable client, including Cursor, OpenAI Codex, Claude Desktop, Claude Code, Windsurf, Cline, Continue, and Zed. Point your client at the registry entry sh.marginal/mcp or the endpoint at https://marginal.sh/mcp and add your API key.
Once connected, the four tools appear to your agent automatically. The docs at https://marginal.sh/docs/ walk through configuration for each client and show example test loops.
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