I just had a single conversation with Claude that ended with an article published to our website and automatically distributed to Twitter and LinkedIn. No tab switching. No copy-pasting. Just a conversation.
Here's exactly how I built it and what it means for AI-powered operations.
The Starting Point
Gadoci Consulting runs on a custom platform that manages our articles, learning tracks, and client portal. We recently built an MCP (Model Context Protocol) connector that lets Claude interact directly with our system. Claude can search articles, create new content, manage prospects, and pull analytics—all through natural conversation.
The MCP connector was the foundation. But content creation is only half the job. Distribution is the other half, and that's where things got interesting.
The Challenge
I wanted to publish an article and share it to social media without leaving my conversation with Claude. The problem: Claude can talk to our platform through MCP, but it can't natively post to Twitter or LinkedIn.
The solution came through n8n, an open-source automation platform that also supports MCP. By connecting n8n to Claude, I could build webhook-triggered workflows that Claude can execute directly.
Building the Twitter Pipeline
The Twitter integration was straightforward. In n8n, I created a workflow with two nodes:
- A Webhook trigger that accepts POST requests with a
contentfield - A Twitter node that posts the content directly
After some debugging with expression syntax (turns out $json.body.content is the right path when MCP sends webhook data), the pipeline worked. Claude can now post to Twitter by calling the n8n workflow.
The LinkedIn Workaround
LinkedIn was trickier. Their API restricts posting to approved developer applications, and the approval process is designed for enterprise integrations, not small consultancies.
The workaround: Zapier. Zapier has pre-built LinkedIn OAuth connections that handle the complexity. So the LinkedIn pipeline became:
n8n Webhook → HTTP Request to Zapier → Zapier posts to LinkedIn
It's an extra hop, but it works reliably. Claude calls n8n, n8n calls Zapier, Zapier posts to LinkedIn. The whole chain executes in under a second.
What the Workflow Looks Like Now
When I want to publish and distribute content, the conversation goes like this:
Me: "Post The Great Compaction article to Twitter and LinkedIn."
Claude: Pulls the article content from our platform via MCP, drafts platform-appropriate copy, executes the n8n workflows for both Twitter and LinkedIn, and confirms the posts are live.
That's it. One message, full distribution.
The Bigger Picture
This isn't about saving time on social media posts. It's a proof of concept for something larger: AI agents that can take real actions across your entire business stack.
The pattern we established here applies everywhere:
- MCP connects Claude to your custom systems
- n8n provides webhook-triggered automation for anything MCP can't reach natively
- Zapier fills gaps where APIs are restrictive or OAuth is complex
Each tool has its strength. MCP excels at deep integration with custom platforms. n8n excels at flexible automation with a visual builder. Zapier excels at pre-built connections to hundreds of services.
Chained together, they give Claude the ability to operate across systems that were never designed to talk to each other.
Practical Considerations
A few things I learned during the build:
Expression syntax matters. n8n's Twitter node was configured with $json.content but the webhook wraps incoming data in a body object. The fix was changing it to $json.body.content. Small detail, frustrating to debug.
Test with real data early. Zapier's webhook needed a test payload to recognize the data structure. Sending a curl request with sample content let Zapier map the fields correctly.
Activate workflows in MCP. n8n workflows have a setting called "Available in MCP" that must be enabled. Without it, Claude can't see or execute the workflow.
What's Next
This pipeline handles text posts. The next iteration will add image generation, so Claude can create featured graphics before distributing. After that, scheduling—queue posts for optimal timing rather than immediate publishing.
The foundation is in place. Now it's about expanding what Claude can do with it.