How We Built a Blog-to-Social Pipeline Using Claude AI in Two Weeks

The problem was simple: we were writing technical blog posts that nobody in the C-suite was reading. They were good posts — detailed, accurate, genuinely useful — but written for engineers. And the social media presence was essentially non-existent. Posting manually on LinkedIn once every few weeks wasn't moving the needle.

So we built a pipeline to solve both problems at once. One trigger. Four outputs. Zero manual effort after setup.

The End State

Here's what the pipeline does every week, automatically:

1. A new technical blog post is published to the site
2. Claude reads it and rewrites it for an executive audience — same facts, no jargon, business-impact framing
3. The executive version is published as a new blog entry on the site automatically
4. Claude generates three platform-specific social posts: LinkedIn (thought leadership format), X (punchy, hook-first), Facebook (community-oriented)
5. Posts are scheduled and published to each company page throughout the week

Total human time after initial setup: zero minutes per week. The pipeline runs on a cron trigger every Monday morning.

The Stack

We kept the tool count tight:

• Claude (claude-sonnet-4-6) — the AI doing the writing and rewriting
• Make.com — the automation orchestrator connecting everything
• Cloudflare Pages — where the site lives (static HTML, fast, cheap)
• LinkedIn API, X API, Facebook Graph API — the publishing endpoints
• Google Sheets — optional approval queue (we added this after the first week)

We also used Claude Code (Anthropic's CLI tool) to write several of the helper scripts that manage the site's blog generation. That meta-layer — using Claude to build the Claude pipeline — saved a significant amount of time during construction.

Step 1: The Trigger

Make.com has a built-in RSS/webhook trigger. We pointed it at the site's RSS feed. When a new post appears, the scenario fires. Simple.

For the weekly cadence, we added a secondary trigger: a Make.com schedule that runs every Monday at 9am. If there's a new post in the feed that hasn't been processed, it picks it up. If we published something mid-week, it processes immediately. Both paths lead to the same next step.

Step 2: Claude Rewrites for Executives

This is where the real work happens. We send the full blog post content to Claude via Make.com's HTTP module, calling the Anthropic API directly.

The prompt took several iterations to get right. The key insight: you can't just say "rewrite this for executives." You need to be specific about what executives care about and what they don't.

Our final prompt structure:

The output is consistently usable with minimal editing. Occasionally a line needs a tweak — that's what the approval queue in Google Sheets is for.

Step 3: Publishing the Executive Version

We use the Cloudflare Pages API to push a new HTML file to the repository. Make.com calls a small Cloudflare Worker that accepts the article content, wraps it in the site's blog post template, and writes the file. Cloudflare Pages automatically deploys on commit.

The blog listing page (blog.html) gets regenerated at the same time with the new entry appended. Total latency from trigger to live page: about 90 seconds.

Step 4: Social Post Generation

Three separate Claude calls — one per platform. Each uses a different system prompt tuned for that platform's format and audience behavior.

LinkedIn prompt guidance: professional tone, 150-250 words, starts with a hook sentence, uses line breaks for readability, ends with a question to drive comments, 3-5 relevant hashtags.

X prompt guidance: under 280 characters for the main tweet, punchy opener, link to the executive blog post, 1-2 hashtags max. We also generate a 3-tweet thread variant for posts where the topic warrants it.

Facebook prompt guidance: conversational, slightly longer than LinkedIn, community-focused framing, ends with a question or invitation to share experiences.

Step 5: The Approval Queue (Optional but Recommended)

For the first month, we ran fully automatic. The quality was good enough that we could — but we added an approval step anyway, not because the AI was wrong, but because it's useful to review what's going out under your name.

Make.com writes the three posts to a Google Sheet with columns for each platform, the post text, a status field, and a publish date. We review and either leave the status as "approved" (default) or change it to "edit needed." Another Make.com scenario polls the sheet every few hours and publishes approved posts on schedule.

If you want a fully hands-off system, skip the sheet and publish directly. Your call.

What It Cost to Build

About two weeks of part-time effort. The components that took the longest: getting the Cloudflare Pages publish step working reliably, and iterating on the Claude prompts until the output was consistently usable without editing. The Make.com orchestration itself was straightforward.

Ongoing API costs are minimal. Claude API calls for a typical 1,500-word blog post cost under $0.05 per run. Make.com operations are well within the free tier limits for this scenario. Social API calls are free.

What We'd Do Differently

We'd add the Google Sheets approval queue from day one instead of retrofitting it. And we'd build a simple dashboard that shows the last 10 posts processed, their status, and what was published where — we're still using the Make.com execution history for that, which isn't ideal.

Can You Build This For Your Business?

Yes — and it doesn't take two weeks anymore because we've already done the hard parts. If you're publishing content regularly and want the social amplification and executive version handled automatically, book a discovery call and we'll scope what this looks like for your specific setup.