A GDPR-conscious email testing platform for AI agents
If you operate in the EU or send to European subscribers, every subject line A/B test you run carries data-protection obligations. marginal is a hosted email MCP server that lets AI agents generate variants, launch tests, and read lift metrics — so your experimentation workflow stays auditable and tight on what data moves where.
Why GDPR matters when you test subject lines
Subject line testing looks harmless, but it touches personal data the moment you map opens and clicks back to identifiable recipients. A GDPR email testing platform should keep that processing scoped, purposeful, and easy to explain to a DPO.
marginal focuses on the experiment layer: it generates variants, runs the managed send, and reports aggregate open/click and within-test lift. Your testing decisions are driven by cohort metrics rather than sprawling individual-level profiling.
- Test results are reported as aggregate open/click and lift, not per-person dossiers
- Clear, single purpose: optimizing subject line performance
- Bearer API key auth (marg_live_...) keeps access scoped to your account
- Hosted endpoint at https://marginal.sh/mcp — no self-managed infrastructure to audit
How the email MCP server fits your stack
marginal is an email MCP server, meaning AI agents talk to it through the Model Context Protocol. Instead of wiring custom scripts, your agent calls four tools to run the full loop: generate_variants, launch_test, get_results, and recommend_next.
Because it's a hosted service reachable at https://marginal.sh/mcp, there's no self-hosted component to patch or secure. That reduces the surface area you have to reason about when documenting data flows for compliance reviews.
- generate_variants — draft subject line candidates
- launch_test — start a managed subject line A/B test
- get_results — pull open/click and within-test lift
- recommend_next — let the agent suggest the next experiment
AI agent email marketing without losing control
AI agent email marketing is fast, but speed without guardrails is how teams accidentally over-collect or over-retain data. marginal keeps the agent's job narrow: propose copy, run the test, read the numbers. Your agent never needs to hold raw recipient lists to make a subject line A/B test decision.
Connect from the client your team already uses — Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, or OpenAI Codex — and let the agent iterate on subject lines while you retain oversight of what gets sent and why.
- Agents act on aggregate lift, not personal identifiers
- Works across major MCP-capable clients
- Each test has a defined start, send, and result-reporting boundary
Start testing within compliance-friendly limits
The free tier includes 100 experiments per month, which is enough to validate your subject line workflow and your internal data-handling process before scaling. Documentation at https://marginal.sh/docs/ covers tool schemas and connection setup.
Pair marginal with your own consent and retention policies, and you have a practical foundation for treating subject line experimentation as a controlled, GDPR-aware activity rather than an afterthought.
- 100 experiments/month on the free tier
- Docs: https://marginal.sh/docs/
- Registry: sh.marginal/mcp
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