npx content-grade · Claude CLI · No API key needed

Your content is competing
against AI-generated noise.
Grade it before you ship it.

ContentGrade is a suite of 6 AI-powered content analysis tools — headline grader, landing page auditor, ad scorer, thread grader, email forge, and audience decoder. Runs on your local Claude CLI. No API key. No cloud. Data stays on your machine.

Install free — npx content-grade → See the 6 tools

~ npx content-grade ./my-post.md

    OVERALL SCORE   72/100  B+   blog post
    ████████████████████████░░░░░░░░░░░░░░░░

    HEADLINE SCORE   61/100
    "Why Most Startups Fail"
    ↳ Too vague — the reader doesn't know if this applies to them.

    DIMENSION BREAKDOWN

    Clarity        80/100  ↳ Well-organized with clear section breaks.
    Engagement     60/100  ↳ Opens strong but drops in paragraph 3.

    TOP IMPROVEMENTS
    ● Make the headline specific: "Why 90% of SaaS Startups Fail at Month 18"
    ● Add one data point in paragraph 2 — "CB Insights found..." beats vague claims

Free · 3 analyses/day per tool · Claude CLI required · Zero data leaves your machine · Open source · MIT license

The problem

AI made content creation instant.
It didn't make it good.

Every competitor, agency, and solo founder is publishing more content than ever. Most of it scores below 30. The teams who measure and improve win on signal-to-noise ratio alone.

Generic doesn't convert

"Learn More About Our Solutions" and "How 847 Founders Cut AWS Costs 40% in 30 Days" are both 10 words. Only one has a dream outcome. Word-counting tools can't see the difference.

No feedback loop

You ship the headline, the ad, the email. A/B tests take weeks. By then the campaign is over. You need instant, framework-based feedback before the content goes live — not after.

SaaS tools own your data

Agencies and teams run confidential client briefs, unreleased product copy, and internal strategic messaging through these tools. Cloud SaaS logs everything. ContentGrade runs on your machine via Claude CLI — nothing leaves your device.

The suite

Six tools.
One content quality system.

Each tool targets a specific content type with the frameworks that actually matter for that format. No generic "AI writing assistant" scores here.

✍️

HeadlineGrader

Score any headline 1–100 across Masterson's Rule of One, Hormozi's Value Equation, Proof + Promise + Plan, and Readability. Includes AI rewrites labeled with the technique applied.

Copywriting frameworks · SERP preview · A/B compare

🔥

PageRoast

Drops your landing page into a conversion audit. Identifies friction points, weak CTAs, missing proof elements, and headline-body mismatch. Brutally honest, actionably specific.

Conversion audit · CTA analysis · Friction scoring

📢

AdScorer

Grade ad copy for Google, Meta, and LinkedIn against platform-specific frameworks. Hook strength, offer clarity, proof density, and CTA quality — scored per dimension with rewrites.

Google · Meta · LinkedIn · Platform-aware scoring

🧵

ThreadGrader

Evaluate Twitter/X threads for hook strength, retention across tweets, value density, and CTA placement. Scores each tweet individually and the thread as a whole.

Hook analysis · Retention scoring · Per-tweet breakdown

📧

EmailForge

Analyze email subject lines and body copy. Open rate prediction, subject line A/B comparison, body flow analysis, and CTA effectiveness. Works for cold outreach, newsletters, and sequences.

Subject line scorer · Open rate prediction · A/B compare

✦

ContentGrade

The master scorer. Paste any content — blog post, case study, sales page, product description. Get a holistic quality score across clarity, specificity, proof density, audience fit, and conversion potential.

All content types · Holistic scoring · 6-dimension analysis

The differentiator

Claude CLI.
Local by default.
Zero cloud dependency.

Every other AI content tool is a SaaS that processes your copy on their servers. ContentGrade runs locally on your machine via Claude CLI — no API key required, no monthly Anthropic bill, no data routing through vendor infrastructure.

For agencies running client work, teams drafting unreleased campaigns, and anyone whose copy is legitimately confidential: this is the only option that's actually private.

Client briefs stay off third-party servers
Unreleased campaign copy isn't indexed anywhere
No vendor data retention policy to audit
Your usage patterns aren't training anyone's model
Runs on Claude CLI — free account, no API billing

ContentGrade vs. Cloud SaaS

Data leaves your infra Never

Vendor data retention None — runs on your machine

API key required No — just Claude CLI

Per-query billing None — free Claude CLI

Agency/team use Safe for client work

Framework transparency Open source, fully auditable

Install npx content-grade

Getting started

From install to first grade
in under 3 minutes.

Install Claude CLI

ContentGrade runs on Claude CLI — install it from claude.ai/code (free), then log in. That's the only prerequisite. No API key, no billing setup.

claude login

Run an instant demo

No file needed. Run npx content-grade and it runs a live demo on built-in sample content so you see real output immediately.

npx content-grade
# or analyze your own file:
npx content-grade ./my-post.md

Launch the full dashboard

Start the web interface to access all 6 tools — HeadlineGrader, PageRoast, AdScorer, ThreadGrader, EmailForge, and AudienceDecoder. Free tier: 3 analyses/day per tool.

npx content-grade start

Pricing

Start free. Upgrade when it pays for itself.

3 free analyses/day per tool. No credit card, no signup, no API key required.

FREE

forever · no signup required

Install free →

✓ 3 analyses/day per tool
✓ All 6 tools (CLI + dashboard)
✓ Full framework-based scoring
✓ AI rewrites with technique labels
✓ Per-dimension score breakdown
✓ HeadlineGrader SERP preview
✓ Runs on Claude CLI — no billing

Common questions

What makes this different from CoSchedule Headline Analyzer or other AI writing tools?

Most headline analyzers count words: uncommon word density, emotional word ratio, common word percentage. These are correlates of quality, not causes. They cannot tell the difference between "Learn More About Our Solutions" (empty) and "How 847 Founders Cut AWS Costs 40% in 30 Days" (specific, proof-backed, credible). Both are 10 words. CoSchedule scores them similarly.

ContentGrade uses causal frameworks grounded in direct response copywriting history — Masterson's Rule of One, Hormozi's Value Equation, Proof + Promise + Plan. These explain why a headline fails and give you a specific fix. The other critical difference: your data never leaves your infrastructure. We don't log your copy, and we don't train on it.

Do I need an Anthropic API key or a paid Claude account?

No API key required. ContentGrade runs on Claude CLI — install it for free from claude.ai/code and log in with a Claude account. A free Claude account is sufficient for the free tier (3 analyses/day per tool).

There's no per-query Anthropic billing. You're not routing calls through the API console or accumulating usage charges. Claude CLI handles authentication locally, which is also why your content never leaves your machine.

Is my content private? What data is stored?

Your content goes from your browser → your local ContentGrade server → Claude CLI running locally. It never touches our servers. The only logs that exist are what your local instance stores and whatever Anthropic logs for your CLI session (subject to their standard privacy policy).

For agencies running client work or teams drafting unreleased campaigns: ContentGrade is safe to use. There's no vendor data retention policy to audit because we never have your data.

Which of the 6 tools should I start with?

Start with HeadlineGrader. Headlines are the highest-leverage piece of copy in any content — they determine whether anyone reads the rest. Run your current hero headline, email subject, or blog title. If it scores below 60, the rewrites will show you exactly what to change and why.

After that: PageRoast for landing pages, AdScorer if you're running paid traffic, EmailForge for cold outreach or newsletters. ContentGrade is the catch-all for everything else.

What AI model does ContentGrade use?

ContentGrade uses whatever Claude model your CLI session is authenticated with. By default this is whatever Claude has set as the default for CLI sessions. Since you're running locally via Claude CLI, the model behavior is determined by Anthropic's standard CLI configuration, not by ContentGrade settings.

Can I use ContentGrade for client work at an agency?

Yes, and this is a primary use case. Because ContentGrade runs locally via Claude CLI, client copy never leaves your machine. There's no SaaS terms of service to worry about, no vendor data retention, and no risk of your client's unreleased campaign copy appearing in someone else's training data.

Pro tier includes higher daily limits, and Team tier includes shared workspace features — share analyses and scoring history across your team without routing content through any third-party service.

Your content is competing against AI-generated noise. Grade it before you ship it.

AI made content creation instant.It didn't make it good.