White Label Email A/B Testing API
If you're building a product that sends email on behalf of users, you probably don't want to ship your own experimentation engine. marginal gives you a white label email A/B testing API you can drive directly or through an email MCP server, so your AI agents and backend can run subject line A/B tests without reinventing the stats.
What a white label testing API actually needs to do
A white label email A/B testing API has to handle the unglamorous parts: splitting recipients across variants, sending each one, attributing opens and clicks, and computing lift in a way you can trust. marginal handles all of that behind a single hosted endpoint, so your app surfaces results under your own branding instead of pointing customers to a third-party dashboard.
- Generate competing subject lines from a prompt or supply your own
- Launch a subject line A/B test against a managed send
- Track opens and clicks per variant automatically
- Read within-test lift metrics to declare a winner
- Stay headless — you build the UI your customers see
Driving it through the email MCP server
The same capabilities are exposed as an email MCP server, which matters if your product is built around AI agent email marketing. Instead of stitching together imperative calls, your agent can call typed tools and reason about the results between steps.
The endpoint is https://marginal.sh/mcp, authenticated with a Bearer API key in the marg_live_ format. Connect from Cursor, Claude Desktop, Claude Code, OpenAI Codex, Windsurf, Cline, Continue, or Zed and your agent gets first-class testing tools.
- generate_variants — draft subject line candidates
- launch_test — start a split and begin sending
- get_results — pull opens, clicks, and lift
- recommend_next — decide the follow-up send
A typical embed flow
Because marginal is hosted, there's no MCP server to deploy or maintain on your side. You hold the API key, your agent or backend calls the tools, and you render the outcome inside your own product. A common loop looks like generating variants, launching the test, polling results once enough data lands, then acting on the recommendation.
- Call generate_variants to seed two or more subject lines
- launch_test to split traffic and send
- Wait for engagement, then get_results for lift
- Use recommend_next to schedule the winning follow-up
Pricing and getting started
The free tier includes 100 experiments per month, which is enough to validate the integration and run real subject line A/B tests before you scale. Full tool schemas, auth details, and examples live in the docs at https://marginal.sh/docs/.
- Free tier: 100 experiments/month
- Hosted endpoint, no self-hosting required
- Bearer API key auth (marg_live_...)
- Docs: 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