Content You Don't Pipeline

Markdown in, published out. Search, versioning, and publishing built in. No CMS required.

AI excels at creating content. Blog posts, documentation, reports, emails. But where does that content go? Without a content pipeline, AI-generated content sits in chat logs instead of reaching users.

SystemPrompt provides a complete content management system. Markdown in, published out.

The Problem

Content management for AI products requires:

  • Ingestion: Parse markdown, extract frontmatter, validate structure
  • Storage: Database schema, versioning, drafts vs published
  • Search: Full-text search, filtering, tagging
  • Rendering: Markdown to HTML, syntax highlighting, table of contents
  • Publishing: Static site generation, URL routing, sitemap
  • API: CRUD operations, webhooks, syndication

Building a CMS is a project unto itself. You wanted to ship AI features, not content infrastructure.

The Solution

SystemPrompt includes a content management system:

Markdown-First

Write content in markdown with YAML frontmatter:

---
title: "My Blog Post"
description: "A brief description"
author: "AI Assistant"
slug: "my-blog-post"
public: true
published_at: "2026-01-27"
---

# My Blog Post

Content goes here...

Automatic Ingestion

Content is ingested from the filesystem:

# Directory structure
services/content/
├── blog/
│   └── my-post.md
├── docs/
│   └── getting-started.md
└── config.yaml
# Publish all content
systemprompt core content publish

Search across all content:

# Search content
systemprompt core content search "MCP authentication"

# Filter by source
systemprompt core content list --source blog --limit 10

Static Site Generation

Content renders to static HTML:

# Generated output
web/dist/
├── index.html
├── blog/
│   └── my-post/index.html
├── sitemap.xml
└── feed.xml

Why This Matters for AI

AI-Generated Content Pipeline

Your AI agent creates content. SystemPrompt publishes it:

# services/agents/blog-writer.yaml
name: blog-writer
skills:
  - research_content
  - content_create
mcp_servers:
  - content-server

The agent writes markdown to services/content/blog/. Run systemprompt core content publish. Content is live.

MCP Content Tools

Expose content operations to MCP clients:

# services/mcp/content-server.yaml
name: content-server
tools:
  - search_content
  - create_content
  - update_content
  - publish_content
oauth:
  required: true
  scopes: ["content:read", "content:write"]

Claude Code can search your content, create new posts, and trigger publishing.

Content as Context

AI agents can use your content as context:

  • Search documentation to answer questions
  • Reference existing posts when writing new ones
  • Maintain consistent voice across content

Versioning and Drafts

Content supports drafts and publishing states:

public: false  # Draft, not published
public: true   # Live, visible to users

Review AI-generated content before publishing. Or configure auto-publish for trusted agents.

What You Skip

Without SystemPrompt With SystemPrompt
CMS evaluation and setup Not needed
Markdown parser Built in
Database schema Built in
Search implementation Built in
Static site generator Built in
Publishing workflow CLI command

Getting Started

Create content in services/content/:

# Create a blog post
cat > services/content/blog/my-first-post.md << 'EOF'
---
title: "My First Post"
description: "Getting started with AI content"
author: "Your Name"
slug: "my-first-post"
kind: "article"
public: true
published_at: "2026-01-27"
---

# My First Post

Hello, world!
EOF

# Publish
systemprompt core content publish

# View at /blog/my-first-post

See the Content Reference for detailed configuration options.