Agent Orchestration for Email Campaigns
Modern email work no longer has to live inside a UI. With agent orchestration for email campaigns, your AI assistant can draft variants, launch tests, read results, and decide the next move — all through marginal's hosted email MCP server.
What agent orchestration means for email
Agent orchestration is the practice of letting an AI agent coordinate a multi-step workflow instead of completing isolated one-off tasks. For email campaigns, that means an agent reasons over the whole loop: generate candidate copy, run a subject line A/B test, measure lift, and choose what to send next.
marginal exposes that loop as four MCP tools — generate_variants, launch_test, get_results, and recommend_next — so an agent can move through each stage with a clear, typed contract rather than brittle scripting.
- generate_variants — produce subject line and copy options to test
- launch_test — start a managed send with tracking enabled
- get_results — pull open/click and within-test lift metrics
- recommend_next — let the agent decide the next experiment
How marginal fits the workflow
marginal is a hosted email MCP server, so there is nothing to deploy or self-host. You connect your MCP-capable client to the endpoint at https://marginal.sh/mcp, authenticate with a Bearer API key, and the tools become available to your agent immediately.
Because the server handles managed sends and open/click tracking, AI agent email marketing stays focused on decisions instead of plumbing. The agent orchestrates the campaign; marginal runs the sends and reports the numbers.
- Hosted endpoint: https://marginal.sh/mcp
- Auth via Bearer API key (marg_live_...)
- Registry id: sh.marginal/mcp
- Free tier: 100 experiments per month
A practical orchestration loop
A typical subject line A/B test orchestrated by an agent reads like a short plan. The agent asks for variants, commits to a test, waits for signal, then acts on the recommendation — closing the loop without you switching tools.
- Ask the agent to generate 3–5 subject line variants for a campaign
- Launch a test and let marginal split delivery and track engagement
- Read within-test lift to see which variant is winning
- Use recommend_next to choose the follow-up send or a new test
Clients and getting started
Any MCP-compatible client can drive this workflow, including Cursor, Claude Desktop, Claude Code, OpenAI Codex, Windsurf, Cline, Continue, and Zed. Once connected, you describe the campaign goal in natural language and the agent maps it to marginal's tools.
Full setup details, tool schemas, and examples live in the docs at https://marginal.sh/docs/ — a good next step once you have an API key.
- Works with popular agent and IDE clients out of the box
- Start on the free tier to validate your orchestration loop
- Reference the docs for tool inputs and result fields
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