Score text readability (Flesch, FK Grade, SMOG, ARI) with suggestions via mcp-readability (npx) — 1 tool

Recipe: Score text readability via mcp-readability

Install & launch:

npm install mcp-readability
# server entry: node node_modules/mcp-readability/dist/index.js
# transport: stdio

1 tool available:

• analyze_readability — computes 4 readability metrics + improvement suggestions

Input schema:

{
  "text": "string (required) — the text to analyze, min ~2 sentences for reliable scores",
  "include_suggestions": "boolean (default true) — returns long-sentence, complex-word, and passive-voice flags"
}

What you get back:

• scores.flesch_reading_ease — 0–100+ scale (higher = easier; >60 is "plain English")
• scores.flesch_kincaid_grade — US grade level (9.9 = 10th grade)
• scores.smog_index — years of education needed
• scores.ari — character-based grade level
• reading_level — human label: "Elementary", "Middle School", "High School", "College", "Graduate"
• audience — target reader: "General public", "High school students", etc.
• stats — wordcount, sentencecount, syllablecount, avgwordspersentence, avgsyllablesper_word
• suggestions[] — actionable fixes: "2 words have 5+ syllables — consider simpler alternatives: readability, significantly"

Tested on 3 text types:

1. Mixed text (3 sentences): Flesch 48.5, FK Grade 9.9, SMOG 10.7, ARI 9.8 → High School
2. Dense academic text (2 sentences, legal jargon): Flesch -78.4, FK Grade 29.4, SMOG 22.7, ARI 29.7 → Graduate — flagged 11 complex words and 1 long sentence
3. Simple text (5 short sentences, all monosyllabic): Flesch 116.8, FK Grade -1.7, SMOG 3, ARI -4.3 → Elementary

Scores are internally consistent and differentiate clearly between text complexity levels. Negative Flesch scores for very dense text and >100 for very simple text match the standard formula behavior.

No external API calls. Pure in-process syllable counting and formula computation. Zero config, zero auth.