Let your AI check its own work

Connect to Claude or ChatGPT in fifteen seconds. No key. No cost.

Connector URL
https://levelup-reader.vercel.app/api/mcp

How it works

A chat with Claude: the user pastes a medical passage about atrial fibrillation and anticoagulation and asks to measure it at a target grade of 6. The tool replies with Flesch-Kincaid 18.0, SMOG 12.0, Gunning Fog 21.2, target 6, difficult medical terms flagged as the main issue, and the hardest sentence at roughly grade 28.8.
It calls the tool and gets real numbers.
The same chat continuing: the user asks Claude to rewrite the passage at grade 6 and keep checking until it measures correctly. Claude's rewritten passage is shown along with a tool result reading Flesch-Kincaid Grade 5.7, which meets the Grade 6 target, alongside SMOG 9.5 and Gunning Fog 9.0.
It writes, checks itself, and revises until it lands.

Add it

  1. Copy the URL above.
  2. Claude: Settings → Connectors → Add custom connector → paste.
  3. ChatGPT: turn on Developer Mode in Settings first, then the same steps.

No OAuth. Leave any login fields blank.

Use it

Paste text, ask for its reading level. You get three formulas, a passive-voice count, and the sentences hurting you most.

Set a target grade and it tells you if you hit it.

Ask it to rewrite something and check its own work. It keeps going until the number is right.

Hand it several pages, sections, or drafts at once, and it will rank them worst-first.

Why it's free

Readability is arithmetic, not AI. No model call, no key, no limit.

It flags disagreement

On 49 of 49 real health pages, Flesch-Kincaid rated the text easier than SMOG and Gunning Fog did. FK is the formula most people audit with. When the three spread apart, we say so instead of handing you the friendliest number.

Add to ClaudeAdd to ChatGPT

ChatGPT needs Developer Mode switched on first.