Tomas Lebedinskas

Posted on Apr 15 • Originally published at godberrystudios.com

How I Built a Content Repurposing MCP Server on Apify (and How to Use It)

#mcp #ai #apify #typescript

The Problem

If you've ever written a blog post and then had to manually repurpose it for Twitter, LinkedIn, Instagram, Facebook, and email — you know the pain. Each platform has different formats, character limits, tone expectations, and engagement patterns.

I was spending 30-45 minutes per article just rewriting the same content for different platforms. Multiply that by 3-4 articles per week, and you're looking at hours of repetitive work that AI should be handling.

The Solution: An MCP Server

I built Content-to-Social MCP — a hosted MCP server that takes any blog post or article URL and returns platform-optimized social media content for all 5 major platforms in a single call.

Here's what it looks like in practice. You give it a URL:

Transform this article into social media posts:
https://example.com/blog/interesting-article

And it returns ready-to-post content for:

Twitter/X — A thread with hook, key points, and CTA
LinkedIn — Professional tone with industry insights
Instagram — Caption with relevant hashtags
Facebook — Conversational, shareable format
Email newsletter — Subject line + body snippet

All optimized for each platform's best practices.

How It Works Under the Hood

The architecture is simple:

URL → Content Extraction (Cheerio) — Fast HTML parsing, no headless browser needed. Extracts title, author, date, and the main article body.
Content → Platform Posts (Claude Sonnet) — The extracted content is sent to Claude with platform-specific prompts that enforce character limits, tone, and formatting rules.
Hosted on Apify Standby — The server runs on Apify's Standby mode, which means it's always warm and ready to respond. No cold starts, no server management, no Docker headaches.

The key technical decisions:

Cheerio over Puppeteer — 10x faster, 10x cheaper. Most blog content is in the initial HTML.
Streamable HTTP over SSE — The MCP protocol recently moved to Streamable HTTP as the standard transport. This plays nicely with Apify's reverse proxy.
Stateless architecture — No sessions, no state. Each request is independent. This makes scaling trivial.

How to Use It (3 Minutes Setup)

Option 1: Claude Desktop

Add this to your Claude Desktop MCP config (claude_desktop_config.json):

{
  "mcpServers": {
    "content-to-social": {
      "url": "https://godberry--content-to-social-mcp.apify.actor/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_APIFY_TOKEN"
      }
    }
  }
}

Replace YOUR_APIFY_TOKEN with your Apify API token (free tier available at apify.com).

Restart Claude Desktop. You now have 3 new tools:

extract-article (free) — Preview what the server extracts from a URL
transform-for-platform ($0.03) — Generate content for a single platform
transform-content ($0.07) — Generate content for all 5 platforms at once

Option 2: Any MCP-Compatible Client

The endpoint works with Cursor, Windsurf, Cline, or any client that supports MCP over Streamable HTTP. Just point it at:

https://godberry--content-to-social-mcp.apify.actor/mcp

Option 3: Direct HTTP

You can also call it directly via HTTP POST if you want to integrate it into your own tools:

curl -X POST https://godberry--content-to-social-mcp.apify.actor/mcp \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"method":"tools/call","params":{"name":"transform-content","arguments":{"url":"https://example.com/article"}}}'

The Business Model

This runs on Apify's Pay-Per-Event pricing:

$0.07 per full transformation (all 5 platforms)
$0.03 per single-platform transformation
Free article extraction (preview before you pay)

No monthly subscription. You pay only when you use it. For context, manually repurposing one article takes 30-45 minutes. At $0.07, this is roughly 1,000x cheaper than your time.

Lessons From Building This

1. Ship the simplest version first. My first version tried to handle edge cases for every possible website. The shipping version just uses Cheerio with sensible defaults. It works on 90%+ of blog posts, and that's enough.

2. Apify Standby mode is underrated for MCP servers. Most MCP server tutorials tell you to self-host with Docker. Apify's Standby mode gives you a persistent URL, auto-scaling, and built-in billing — without managing any infrastructure.

3. The MCP ecosystem is moving fast. When I started building, SSE was the standard transport. By the time I deployed, the ecosystem had shifted to Streamable HTTP. Stay close to the spec.

4. Content generation MCPs are different from tool MCPs. Most MCP servers are "tool" servers — they connect AI to external APIs (Slack, GitHub, databases). Content generation servers create new content from existing content. The market for this is still wide open.

Try It

Apify Store: https://apify.com/godberry/content-to-social-mcp
MCP Endpoint: https://godberry--content-to-social-mcp.apify.actor/mcp
GitHub: https://github.com/Lebedinskas/content-to-social-mcp-server

If you're a content creator or marketer who uses Claude Desktop, give it a try and let me know what you think. I'm actively building based on user feedback.

I write about MCP servers, web scraping, and automation at godberrystudios.com

DEV Community