marginal: a Brevo alternative built for AI agents
Brevo is a capable ESP, but it was designed for humans clicking through a dashboard. marginal is an email MCP server your agent calls directly — generate variants, launch tests, and read lift metrics without leaving the chat.
Why agents need an MCP email server, not a dashboard
Brevo exposes email campaigns through a web UI and a REST API. That works when a person is driving, but an AI agent has to translate intent into HTTP calls, manage pagination, and stitch together reporting endpoints. marginal removes that friction by speaking the Model Context Protocol natively, so the agent sees a small set of purpose-built tools instead of a sprawling API surface.
If you're evaluating a Brevo alternative for AI agents, the question is less about feature parity and more about how cleanly the agent can reason over the workflow. marginal's tools map one-to-one to the steps of an experiment, which keeps the agent's plan short and predictable.
- generate_variants — draft subject line options for a campaign
- launch_test — start a managed send with an A/B split
- get_results — pull open/click tracking and within-test lift
- recommend_next — get the agent's next action based on results
Subject line A/B testing without the manual setup
Running a subject line A/B test in a traditional ESP means configuring segments, picking a sample size, and waiting for a report you read by eye. With marginal, the agent calls launch_test and gets back structured results — open rate, click rate, and within-test lift between variants — that it can act on immediately.
Because the metrics come back as data rather than a chart, AI agent email marketing loops can close on their own: test, measure, recommend, repeat.
- Generate multiple subject lines from a single prompt
- Launch a managed send with the split handled for you
- Read open/click tracking and lift per variant
- Let recommend_next pick the winning direction
Hosted, with no server to run
marginal is a hosted MCP email server. There's nothing to deploy or self-host — you point your client at https://marginal.sh/mcp and authenticate with a Bearer API key (marg_live_...). That contrasts with stitching together a Brevo API key, your own sending logic, and a custom reporting layer.
The free tier covers 100 experiments per month, which is enough to wire AI agent email into a project and confirm the workflow before scaling. Full setup details live in the docs at https://marginal.sh/docs/.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key — no dashboard juggling
- Free tier: 100 experiments/month
- Registry: sh.marginal/mcp
Works with the agents you already use
Whether your team lives in Cursor, Claude Code, or another MCP client, marginal connects without custom glue code. The same email MCP server tools are available across every supported client, so an experiment authored in one environment behaves identically in another.
If Brevo is your current ESP and you're moving email into agent-driven workflows, marginal slots in as the layer your agents talk to for subject line A/B tests and managed sends.
- Cursor, OpenAI Codex, Claude Desktop, Claude Code
- Windsurf, Cline, Continue, Zed
- One consistent tool set across every client
- Same results format everywhere for reliable agent logic
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