Spam Score & Subject Line Checking via API
Static spam-score checkers guess at whether a subject line will land in the inbox. marginal takes a different angle: it lets your AI agent test subject lines against real recipients and measure what actually performs — open rates, clicks, and within-test lift — all through an API built for agents.
Why a spam score subject line checker API isn't enough
A classic spam score subject line checker API runs heuristics — flagged words, ALL CAPS, excessive punctuation — and returns a probability that a message gets filtered. That's useful as a first pass, but it can't tell you whether a subject line will actually be opened. Two subject lines can both pass a spam check and still perform very differently.
marginal complements static scoring with empirical results. Instead of predicting deliverability in isolation, it measures real engagement so you can optimize for opens and clicks, not just clean spam scores.
- Static checkers estimate filter risk; marginal measures real recipient behavior
- Open and click tracking shows which phrasing earns attention
- Within-test lift quantifies how much a winning subject beats the baseline
How marginal tests subject lines
marginal is a hosted email MCP server for AI agents. You generate candidate subject lines, launch a subject line A/B test, send to your list, and read back the results — all from a small set of MCP tools. There's no infrastructure to run; the server is hosted and you authenticate with a Bearer API key.
An agent loop typically calls these tools in sequence to find and ship the strongest subject line.
- generate_variants — produce subject line candidates to test
- launch_test — start a managed A/B send across variants
- get_results — pull open/click metrics and within-test lift
- recommend_next — get the suggested winner or next experiment
Pairing spam scoring with real A/B testing
A practical workflow is to keep your existing spam score subject line checker API as a pre-flight gate, then let marginal pick the actual winner. Filter out variants that score poorly for deliverability, and route the survivors into a live subject line A/B test to decide based on engagement.
This gives AI agent email marketing workflows the best of both: you avoid obvious spam triggers and you optimize for the metric that matters — whether people open and click.
- Pre-screen variants with your spam checker to reduce filter risk
- Send qualifying variants through marginal's launch_test
- Promote the variant with the highest measured lift
- Free tier includes 100 experiments per month to get started
Connect your AI agent
marginal connects to MCP-compatible clients including Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Zed, and OpenAI Codex. Point your client at the endpoint at https://marginal.sh/mcp and authenticate with your API key to start generating, testing, and scoring subject lines from inside your agent.
See the docs at https://marginal.sh/docs/ for tool schemas and example agent loops.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Works across major MCP clients used by developers and growth engineers
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