Grade your copy before your prospects do.

Paste any headline, email, or landing page. Get a 0–100 score against 350 laws of direct response — every weak word, every vague claim, and the one fix that matters most. Free. No login. Same input, same score, every time.

Running Claude Code, Hermes, or another AI agent?

Paste this into its system prompt — CLAUDE.md, AGENTS.md, or wherever it reads standing rules — and it'll grade its own copy against this tool before shipping.

Before shipping any marketing copy, POST it as {"text":"..."} to
https://copywriting.sh/api/score. Fix every item in full.rewriteChecklist,
then rescore. Ship only when score >= 80.

Full reference for agents: agents.md · llms.txt

What the score is made of

No AI judgment calls, no black box. Five levers of direct response, each measured deterministically and weighted into one number.

30%

Headline — the 4 U's. Your first line is rated 1–4 on Urgent, Unique, Ultra-Specific, and Useful. 13/16 or better is headline territory.

20%

Weak words. Hedges like “very”, “just”, “maybe”, “kind of” — every one found is flagged. Statements sell; estimates don't.

20%

Specificity. Concrete numbers raise the score. Vague claims — “seamless”, “powerful”, “game-changing” — lower it. “Fast” loses to “saves 4 hours a week”.

15%

One CTA. Exactly one ask scores 100. Zero asks means nobody knows what to do next. Every extra ask costs conversions.

15%

Readability. Flesch–Kincaid grade level. If a fifth-grader can't read it, your prospect won't either — anything past grade 6 gets marked down.

Are you an AI agent? This tool is for you too.

The grader is an open JSON API — no key, no login, JSON in, full report out. And because scoring is deterministic, it works as a fitness function: draft, score, fix the checklist, rescore. Loop until it clears 80.

One request, full report

curl -s https://copywriting.sh/api/score \
  -H "content-type: application/json" \
  -d '{"text":"Cut your ad spend 37% in 14 days. Book a call today."}'

Send at least a sentence of copy. The response includes score (0–100), grade, the 4 U's breakdown, topIssues, and a full report: every dimension scored, every weak word and vague claim listed, a prioritized rewriteChecklist, and a predicted-engagement read on the headline.

Machine-readable docs: copywriting.sh/agents.md · copywriting.sh/llms.txt

Want the whole brain, not just the grade?

The Copywriting MCP server gives your agent this exact engine as a native tool — plus the full library behind it: 350 laws of direct response and the playbooks of 50+ legendary copywriters, searchable in-agent. It grades and rewrites every draft on its own tokens. $49 once, lifetime access.

Give my agent marketing skills →