Email Workflow Automation Agents
Email workflow automation agents turn a chat prompt into a measured experiment: generate variants, launch the test, and read back lift. marginal exposes those steps as MCP tools so your agent can run the whole loop without leaving the editor.
What an email automation agent actually does
Most teams think of email automation as drip sequences and timed triggers. An agent-driven workflow is different — it makes decisions about content and tests them against real opens and clicks, then acts on the result.
With marginal, the agent calls a small set of tools to move a campaign from idea to verdict. Each tool is a discrete, auditable step, so you can watch the agent reason through the workflow instead of trusting a black box.
- generate_variants — produce candidate subject lines from a brief
- launch_test — start a managed subject line A/B test with a real send
- get_results — pull open/click tracking and within-test lift metrics
- recommend_next — get the suggested winner or next experiment
Why route email through an MCP server
An email MCP server gives AI agents a typed, predictable interface to sending and measurement. Rather than gluing together API calls and tracking pixels yourself, the agent speaks MCP and marginal handles the send, the tracking, and the statistics.
Because marginal is hosted, there's nothing to deploy or maintain. You connect to the endpoint, authenticate with a Bearer API key, and your AI agent email marketing workflow is live.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Registry id: sh.marginal/mcp
- Free tier: 100 experiments per month
A typical subject line A/B test loop
Here is how an email workflow automation agent runs a subject line A/B test end to end. The agent only needs a short brief — the audience and the message — to drive the rest.
The loop is intentionally tight so you can iterate quickly: generate, launch, measure, decide. Lift is reported within the test, which keeps the agent's recommendation grounded in observed behavior rather than guesswork.
- Agent calls generate_variants to draft 3–5 subject lines
- launch_test sends the variants to a sample and tracks engagement
- get_results returns opens, clicks, and lift per variant
- recommend_next surfaces the winner or proposes a follow-up test
Works with the agents you already use
marginal connects through standard MCP, so the same email workflow automation agents run across editors and assistants. Configure the server once and your tools are available wherever you work.
Supported clients include Cursor, OpenAI Codex, Claude Desktop, Claude Code, Windsurf, Cline, Continue, and Zed. See the docs at https://marginal.sh/docs/ to wire up your client and start your first experiment.
- Cursor, Claude Desktop, and Claude Code
- Windsurf, Cline, Continue, and Zed
- OpenAI Codex
- Docs and setup: 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