Check a LinkedIn draft from inside your AI assistant

Set up

1. In your assistant, add a custom connector or remote MCP server and paste the URL above.
2. Choose no authentication. The server is read-only and needs no account.
3. Paste a LinkedIn draft and ask for it to be checked.

{
  "mcpServers": {
    "linkedin-post-checker": {
      "type": "http",
      "url": "https://q5g4s4bqfitbvcss7a5cnakloe0ynsmd.lambda-url.us-east-1.on.aws/"
    }
  }
}

Tools

• check_linkedin_postScores a draft from 0 to 100 and lists each finding with the quoted text and what to change.
• list_post_rulesLists the 15 rules, the points each deducts and a page explaining each pattern.

The 15 rules

Every draft starts at 100. Each rule takes off 9 points per unit of weight (1, 2 or 3) and is reported once. Drafts under 40 words also get a note, with no points deducted, because the rhythm checks need more text.

• Cliche phrasing
• Figures with no source
• Vague result claim
• Reads like an invented story
• Ends on a question to the reader
• Summary ending
• Universal lesson ending
• Hashtag block
• Markdown asterisks
• Heavy em dash use
• Not just X, it is Y
• Three-item lists
• Enumerated lessons
• Flat rhythm
• Uniform paragraphs

Examples

• Check this LinkedIn draft for AI-sounding patterns: In today’s fast-paced world, founders can’t afford to ignore LinkedIn. Here’s the truth: 87% of buyers research on LinkedIn before they talk to sales. What do you think?

  A score out of 100 with three findings: cliche phrasing, a figure with no source, and a closing question, each with the quoted text and a fix.

• Which rules does the LinkedIn post checker use, and how many points does each take off?

  The 15 rules with their point deductions and a link to the page that explains each pattern.

• Here is my post. Fix whatever the checker flags, then check it again.

  The assistant checks the draft, rewrites the flagged lines, and checks the rewrite so you can compare the two scores.

Limits

• It is rule-based. It matches fixed phrases, patterns and counts, does not use a model, and does not judge meaning.
• It can miss patterns it has no rule for, and it can flag text a person wrote. A true story can be flagged as an invented one.
• English only. Drafts are limited to 20,000 characters.
• The score counts patterns readers associate with generated text. It is not a probability that a model wrote the post.

Privacy and support

The server checks the draft in memory and returns the result. It does not store drafts or sign you in. See the privacy policy and terms. For help, use the contact page. The same rules run in the free browser checker at klype.ai/check, and the open source package is MIT licensed.