Claude Blog Manager: The Conversational CMS

The Project

https://github.com/danielrosehill/Claude-Blog-Manager

Headless CMS platforms offer enormous flexibility for content management, but they come with a tradeoff: the writing experience is often awkward, fragmented, or overly technical. Writing in an IDE does not feel natural. Navigating complex admin interfaces to publish a blog post feels like overkill. This workspace pattern proposes a different approach: what if you could just chat with your CMS?

The Pattern

The Claude Blog Manager combines three things: Claude Code as the conversational interface, a headless CMS (like Contentful) as the content backend, and MCP as the bridge between them. You write your content in a local drafting directory, and then ask Claude to handle the API work needed to push it to your frontend. No context switching to a web admin panel, no wrestling with rich text editors.

The workspace starts with an /init-workspace slash command that configures your blog details and updates the CLAUDE.md system prompt. From there, Claude acts as an intelligent writing assistant that can proofread, add subheadings, insert images with proper relative paths, and handle the entire publication pipeline.

A Cautious Approach to AI Writing

The README is refreshingly honest about AI writing. The author writes by hand and is not interested in generating streams of mediocre content at the push of a button. Instead, Claude serves as an intelligent proofreader that catches issues a spell checker misses, a tool for identifying logical breaks in text and suggesting subheading placements, and an automation layer for tedious formatting tasks like embedding images and writing markdown paths.

The workspace recommends creating one instance per blog you manage, keeping parameters tightly defined and tasks modular. This is a pattern designed for writers who want AI to handle the publishing plumbing while staying firmly in control of the actual writing.