When you are building for scale, technical debt eventually knocks on the door, this weekend, I answered. Here is what shipped:
Sub-Second Search & Pagination Fixes
Thanks to community feedback, I prioritized the "slow search" issue. Digging into the metrics, I found the culprit: classic N+1 query problems that were dragging down the /plugins list response times.
I optimized the underlying queries and eager-loaded the necessary relationships.
Result: Search and pagination now load in under 1 second. Check here to se how i built the site for that
The Lesson: Even experienced developers miss N+1 queries sometimes. This reinforces why optimization isn't just "nice to have" it is a core feature that requires constant observability. When you can't measure it, you can't fix it.
New "Vendor" & "Maintainer" Pages
I have started rolling out dedicated pages for plugin vendors and maintainers.
Current State: Basic listings are live, with clean profiles for major vendors like Spatie, BeyondCode, and others.
- Total downloads across all packages
- Average health score of maintained packages
- Update frequency and responsiveness
- Community trust indicators
What specific data points would you want to see on a Maintainer's profile to judge the quality of their work? Let me know in the comments your feedback drives this roadmap.
With these new pages my site reached the 10k pages count, are you wondering how I host it to allow for such a size? Check my stack in this dedicated article.
Building the "Brain" (MCP Server)
This is the most exciting update. I am actively building a Model Context Protocol (MCP) server for LaraPlugins.
The Vision: Enable AI agents (like Claude or Cursor) to query my directory directly. Instead of hallucinating packages or using outdated packages due to their cut date, your AI editor could ask:
- "What's the best PDF generation package for Laravel?"
- "Show me the most actively maintained charting libraries"
- "Find packages that solve authentication with minimal overhead"
The AI agent would query LaraPlugins in real-time, pulling verified, healthy packages with actual health scores containing "pdf generation" or pdf. No more guessing. No more outdated recommendations. The agent would be then able to choose a plugin based on their health status, php and laravel compatibility, maintenance status and more.
Status: The server is live, but search integration is still a Work In Progress. I am currently refining the caching strategy, since Cloudflare cannot easily cache these dynamic API hits, to ensure stability and performance before opening it to a broader audience.
Why This Matters: Right now, AI models hallucinate package recommendations because they were trained on static data. By providing a real-time MCP interface, I hope to solve a real problem: How do AI agents find the healthiest, most actively maintained Laravel packages?
Quality of Life Updates
- Improved version compatibility badges on plugin cards (clearer visual)
- Refined the "Health Score" algorithm (still tuning this, signals are hard!)
- Added breadcrumb navigation to vendors and maintainers for improved discoverability
Next Steps
Tomorrow is for polishing. I will be finalizing the MCP search integration and ensuring the new Vendor index pages are SEO-ready for organic discovery.
For you: Give the new search a spin and tell me if it feels snappier. Does the sub-second speed make a difference in how you explore packages? As always, your feedback drives this roadmap.
What's coming next: In February, I am building curated collections ("Featured: Filament Ecosystem", "Top 10 Database Packages") and opening early access to the MCP server for beta testers. If you want to be notified when the MCP is ready, drop a comment or reach out in any way.
Until then, happy package hunting.