MCP Prompts

Overview

The MisarBlog MCP server exposes 6 AI writing prompts that generate content using your existing articles as context. Prompts are accessible via MCP clients like Claude Desktop, Claude Code, and any MCP-compatible tool.

Available prompts

draft_article

Fetches your recent article titles as style reference and generates a full Markdown article.

Arguments

Claude Desktop / Code usage

Use the MisarBlog draft_article prompt with topic="Next.js App Router caching strategies"

SDK usage

const client = new MisarBlogMcpClient({ apiKey: "mbk_your_key" });
await client.initialize();
const draft = await client.draftArticle({
  topic: "Next.js App Router caching strategies",
  tone: "technical",
  wordCount: 1200,
});
console.log(draft.messages[1].content.text);

improve_seo

Fetches the content of an existing article, scores its SEO quality, and returns a rewritten version with suggestions.

Arguments

Output

The assistant message includes:

• SEO score (0–100)
• Top 3 weaknesses
• Rewritten title and meta description
• Rewritten introduction paragraph
• Keyword recommendations

SDK usage

const seo = await client.improveSeo({ slug: "my-first-post" });

write_newsletter

Fetches your N most recently published articles and drafts a newsletter issue with curated summaries.

Arguments

SDK usage

const newsletter = await client.writeNewsletter({ articleCount: 5 });

article_ideas

Fetches your top articles by view count, then generates 10 related content ideas in the same niche.

Arguments

SDK usage

const ideas = await client.articleIdeas({ niche: "developer productivity" });

social_caption

Fetches an article's title, excerpt, and canonical URL, then writes captions for Twitter, LinkedIn, and Threads.

Arguments

Output format

The assistant message returns three captions clearly labeled:

• Twitter / X (≤280 characters, includes CTA)
• LinkedIn (professional tone, 2–3 sentences)
• Threads (conversational, hook-first)

SDK usage

const captions = await client.socialCaption({ slug: "my-article" });

optimize_headline

Generates 5 A/B-testable headline variants for an article, each using a different headline strategy.

Arguments

Output format

5 headline variants, each labeled with its strategy:

1. Curiosity gap — creates intrigue
2. How-to — promises a solution
3. List — numbered outcome
4. Specific number — data-driven credibility
5. Bold claim — contrarian or surprising

SDK usage

const headlines = await client.optimizeHeadline({
  title: "How to Write Better Articles",
});
console.log(headlines.messages[1].content.text);

Prompt response format

All prompts return the MCP GetPromptResult structure:

{
  messages: [
    {
      role: "user",
      content: {
        type: "text",
        text: "// Prompt text with fetched context"
      }
    },
    {
      role: "assistant",
      content: {
        type: "text",
        text: "// AI-generated content"
      }
    }
  ]
}

Claude Desktop configuration

See MCP Setup to add the MisarBlog MCP server to Claude Desktop. Once connected, invoke prompts naturally:

"Use the MisarBlog article_ideas prompt for developer productivity content"

"Run the MisarBlog optimize_headline prompt on my article 'How to Write Better Code'"