Campaign Monitor Alternative for AI Agents
If you want your AI agent to own subject line testing instead of clicking through a campaign dashboard, marginal is a Campaign Monitor alternative built around the Model Context Protocol. Agents call tools directly to generate variants, launch tests, and read lift metrics.
Why agents need a different tool than Campaign Monitor
Traditional email platforms like Campaign Monitor are designed for humans navigating a UI. That works fine for marketers, but it's friction when an autonomous agent is doing the work. There's no clean, structured surface for an LLM to call.
marginal flips that model. It exposes an email MCP server at https://marginal.sh/mcp, so AI agent email marketing workflows happen through typed tool calls — no scraping, no brittle automation against a web interface.
- Structured MCP tools instead of UI clicks or undocumented internal APIs
- Hosted service — there's nothing to deploy or self-host
- Auth is a single Bearer API key (marg_live_...)
- Free tier includes 100 experiments per month to evaluate it
What a Campaign Monitor alternative for agents should do
The core loop for any agent-driven email program is the same: draft options, ship a test, measure, and decide what to send next. marginal maps each step to a dedicated tool so an agent can run the whole cycle without a person in the middle.
- generate_variants — produce subject line candidates to test
- launch_test — start a subject line A/B test with managed sends
- get_results — pull open/click tracking and within-test lift metrics
- recommend_next — get the suggested winner or next move
Subject line A/B testing without the dashboard
Subject lines are where most email lift lives, and they're exactly the kind of repetitive, data-driven task agents handle well. With marginal, an agent can spin up a subject line A/B test, wait for engagement data, and recommend the strongest variant — all programmatically.
Because tracking and lift calculation are handled inside each test, your agent doesn't need to reconstruct stats from raw event logs. It reads the result and acts.
Connecting it to your stack
marginal works with the MCP clients developers already use, so adding it to an existing agent setup is mostly a config change. Once connected, the same tools are available across editors and assistants.
Supported clients include Cursor, OpenAI Codex, Claude Desktop, Claude Code, Windsurf, Cline, Continue, and Zed. See the docs at https://marginal.sh/docs/ for setup details and tool references.
- Point your client at the endpoint: https://marginal.sh/mcp
- Authenticate with your marg_live_ API key
- Call generate_variants then launch_test to start your first experiment
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