a2zwebltd / laravel-content-manager

v1.0.3 2026-09-17 07:40 UTC

Requires

• php: ^8.2
• laravel/framework: ^11.45.3|^12.41.1|^13.0
• laravel/mcp: ^1.0
• league/commonmark: ^2.4
• spatie/laravel-medialibrary: ^11.0

Requires (Dev)

• laravel/pint: ^1.25
• orchestra/testbench: ^9.0|^10.0|^11.0
• pestphp/pest: ^3.0|^4.0
• pestphp/pest-plugin-laravel: ^3.0|^4.0
• spatie/laravel-feed: ^4.5

Suggests

• laravel/ai: Enables the AI draft pipeline (content:generate-drafts and the posts.draft MCP tool).
• laravel/nova: Auto-registers Nova resources (BlogPost, ContentCategory, Tag, Page, Faq, Chunk) when present.
• spatie/laravel-feed: Serves the blog as an Atom/RSS feed via A2ZWeb\ContentManager\Feeds\FeedableBlogPost.
• spatie/laravel-responsecache: Public HTTP cache is flushed automatically whenever content changes.

MIT

• A2Z WEB
• Dawid Makowski

• blog
• cms
• markdown
• nova
• content
• laravel
• ai
• mcp

Published on Packagist.org on 2026-09-17 07:40 UTC

README

A portable content engine for Laravel: a markdown blog with categories and tags, database-backed CMS pages, FAQ entries, reusable content chunks, media-library images and Nova admin — plus a built-in MCP server, so AI agents can run the whole archive remotely with nothing but an API key.

Designed to drop into any Laravel app: every table name, model, route and view is overridable, so it fits an existing blog as readily as an empty project.

Integrating it into an app? INTEGRATION.md is a step-by-step guide written for an AI coding agent. Point one at it — "integrate a2zwebltd/laravel-content-manager into this app following its INTEGRATION.md" — and it covers install, configuration, routes, views, adopting an app that already has a blog (including the morph-type backfill that silently drops every tag if it is skipped), the MCP server, and a verification checklist.

Raw URL for an agent that needs to fetch it: https://raw.githubusercontent.com/a2zwebltd/laravel-content-manager/main/INTEGRATION.md

Connecting an agent to a site that already runs this package? MCP.md is the usage reference: every tool and its arguments, the read-before-write conventions, the traps (soft-deleted slugs keep their URL, chunks render live, force is permanent) and what each error means. Raw URL: https://raw.githubusercontent.com/a2zwebltd/laravel-content-manager/main/MCP.md

Requirements

• PHP 8.2+
• Laravel 11.45.3, 12.41.1 or 13
• spatie/laravel-medialibrary (required — featured images)
• laravel/mcp (required — the MCP server)
• Optional: laravel/nova (admin resources), laravel/ai (AI draft pipeline), spatie/laravel-feed (Atom/RSS), spatie/laravel-responsecache (auto-flushed on every content change)

Installation

composer require a2zwebltd/laravel-content-manager
php artisan vendor:publish --tag=content-manager-config
php artisan migrate

That is the whole install: /blog, /blog/tags, /blog/tag/{slug}, /blog/category/{slug}, /blog/{slug} and the whitelisted page route are live, with the package's own views.

Publish the views to restyle them:

php artisan vendor:publish --tag=content-manager-views

Features

• Blog — markdown posts with intro, subtitle, SEO meta, promotion flag, soft deletes, scheduled publishing, LIKE search, related posts (tags → categories → latest) and chronological neighbours.
• Taxonomy — categories and polymorphic tags, both with their own landing pages and descriptions.
• Pages — database-backed CMS pages whose route binding is scoped to published rows, so a draft 404s rather than leaking.
• FAQ — ordered entries, optionally grouped so each page renders its own set.
• Chunks — named snippets a template drops in by code, cached and invalidated on save.
• Images — media-library collections (main, small) with full_size, preview and thumb conversions; the Nova upload writes through the media library, so the admin and the front end agree on one file.
• MCP server — 22 tools, two resources and a prompt, behind a static API key.
• AI drafts — optional content:generate-drafts writes the backlog into unpublished posts for a human to review.
• Sitemap and llms.txt helpers — ContentUrls and ContentIndex hand the host plain arrays, so the package never has to depend on a sitemap package.

The MCP server

Set a key and the endpoint appears; leave it unset and the route is never registered at all.

CONTENT_MCP_API_KEY="<generate-a-long-random-string>"

# or several, each with its own abilities:
CONTENT_MCP_API_KEYS="writer:<key-one>:read|write,reader:<key-two>:read"

php artisan content:mcp-status     # is it registered, and what may each key do?
php artisan mcp:inspector mcp/content

Point an agent at it:

claude mcp add --transport http my-site-content https://example.com/mcp/content \
  --header "Authorization: Bearer <your-key>"

Full usage reference for a connected agent — arguments, workflows, conventions and error semantics: MCP.md.

Tools

Area	Tools
Posts	list-posts, get-post, create-post, update-post, publish-post, unpublish-post, delete-post, restore-post, set-post-image
Taxonomy	list-categories, create-category, list-tags, create-tag
Pages	list-pages, get-page, upsert-page
FAQ	list-faqs, upsert-faq, delete-faq
Chunks	list-chunks, upsert-chunk
Meta	content-stats

Resources: content://guidelines (house style, built from your config) and content://taxonomy (the category and tag slugs that already exist). Prompt: draft-post.

A read-only key that calls a write tool gets a plain explanation back, not a crash. Every mutation is logged with the name of the key that made it.

Routes

Method	URI	Name
GET	/blog	blog.index
GET	/blog/tags	blog.tags
GET	/blog/tag/{tag}	blog.tag
GET	/blog/category/{contentCategory}	blog.category
GET	/blog/{blogPost}	blog.show
GET	/{page} (whitelisted slugs)	pages.show
POST	/mcp/content	—

Configuration

Everything lives in config/content-manager.php, which is closure-free and safe to config:cache.

• models — swap in your own subclass of any model; relations, controllers, Nova and the MCP tools all resolve through this map.
• morph_map / legacy_morph_types — tag pivots and media rows store the morph type as a string. The package aliases it (blog_post, content_page) and ships a migration that rewrites legacy class names, so adopting the package in an app that already had its own App\Models\BlogPost does not silently orphan every tag and image.
• tables — prefix or rename any table. tags and pages are the ones most likely to collide with something you already own.
• routes — set enabled => false to keep your own routes (and your own URLs, middleware and ordering) while still using the package's controllers.
• views — point any view at one of your own Blade files. The view data contract is frozen and covered by tests: posts, search, categories, tags, totalPosts, post, related, previousPost, nextPost, tag, category, groupedTags, page.
• events — ContentSaved / ContentDeleted fire on every change. The bundled listener clears the response cache; on_change takes invokable class-strings for anything else (rebuilding a sitemap, purging a CDN).
• editorial — the house style served to agents and used to build the AI prompt.
• ai — provider, model and the topic backlog for content:generate-drafts.

Adopting it in an app that already has these tables

Full procedure: INTEGRATION.md §6.

The migrations are guarded with Schema::hasTable(), so they record themselves and change nothing where the tables already exist. Delete your own create-migrations, keep their rows in the migrations table, and run the morph normalisation the package ships. Verify with:

SELECT taggable_type, COUNT(*) FROM taggables GROUP BY 1;   -- expect only the aliases
SELECT model_type, COUNT(*) FROM media GROUP BY 1;

Feed

// config/feed.php
'items' => [A2ZWeb\ContentManager\Feeds\FeedableBlogPost::class, 'getFeedItems'],

Then set content-manager.models.blog_post to the same class so the feed, the morph alias and Nova all agree on one model.

Testing

composer test
composer lint

Security Vulnerabilities

Please report security issues to [e-mail].

License

MIT. See LICENSE.

Credits

• A2Z WEB
• Dawid Makowski