Is Flight PHP Dead in the AI Era?

The PHP ecosystem moves fast, and AI-assisted coding has accelerated everything. If you’ve been around for a while, you’ve probably seen the chatter: “Is Flight PHP dead?” With LLMs favoring popular stacks and some developers hitting friction when prompting for Flight patterns, it’s a fair question. But let’s cut through the noise with real data and hands-on experience.

Flight isn’t dead. It’s quietly positioned as one of the smartest choices for developers who value simplicity, speed, and delightful workflows—especially when pairing it with modern AI tools.

The Data Doesn’t Lie: Flight Is Seeing Real Momentum

Recent package statistics for flightphp/core show installs hitting all-time highs, with a noticeable uptick in activity.

This aligns with broader PHP trends: a spike in PHP 7.4 usage.

Many teams and hosts are still managing legacy codebases or prioritizing stability during migrations. Flight’s long-standing PHP 7.4+ support and minimal footprint make it a natural fit for these environments.

Flight’s core remains tiny—easily fitting inside most AI context windows (around 51k tokens including tests). That’s a massive advantage over heavier frameworks where truncation leads to hallucinations. Backwards compatibility has been a core tenet since 2011, meaning older examples and patterns still work reliably in v3. This stability is gold for AI tools trained on years of consistent Flight code.

Addressing the AI Friction Head-On

I’ll be transparent—I’ve felt the pain myself. In recent experiments building an AI-focused Flight skeleton, models (even strong ones like Claude or Minimax) sometimes fought against the framework’s clean, no-magic style. They’d push Laravel conventions or assume heavier structures because that’s what dominates training data.

Reframing prompts around Flight’s principles helped, but it highlighted a real gap. That led me to explore alternatives and ultimately create the slim-ai-skeleton—a straightforward setup using Slim and complementary libraries that felt more intuitive for current LLMs. If Flight’s patterns aren’t clicking with your AI workflow right now, Slim is a solid, lightweight option worth checking out. No shame in using the right tool for the job.

That said, Flight has been catching up fast with native AI support that plays to its strengths.

Flight’s Built-In AI Advantages

The official docs now feature a dedicated AI & Developer Experience section. Here’s what makes it stand out:

• Runway CLI: Flight’s command-line tool includes ai:init (to connect providers like OpenAI, Grok, or Claude) and ai:generate-instructions. This crafts project-specific guidance for GitHub Copilot, Cursor, Windsurf, and more—covering your DB choice, templating (Latte shines here), routing style, and security conventions.

• Official Skeleton: composer create-project flightphp/skeleton my-app gives you a ready-to-go structure with pre-configured AI instruction files. Serve it with composer start and you’re off.

• FlightPHP MCP Server: This is a game-changer. The publicly hosted MCP server at https://mcp.flightphp.com/mcp gives any MCP-compatible AI assistant (Claude, Cursor, etc.) live, structured access to the full Flight documentation—no setup or keys needed. Point your agent at it and watch context quality improve dramatically. I actually discovered with Claude Code that when working with Flight, Claude would request the docs web page, but then the web_fetch tool would summarize it and present it back to Claude code....only it was missing HALF of the documentation! The MCP server ensures you get the right documentation into the context window of your conversation!

These tools reduce the “fighting the AI” problem by feeding it precise, up-to-date Flight knowledge.

Why Simplicity Wins with AI (and Humans)

Flight’s ethos—Simple, Fast, Delightful—aligns perfectly with effective AI prompting. Less magic means clearer intent, fewer surprises, and code that’s easier for both humans and models to reason about. Route grouping, middleware, dependency injection, and plugins like ActiveRecord or Latte keep things lightweight without boilerplate overload.

For new users, this means you can understand the entire request lifecycle quickly. For veterans, it means maintainable apps that scale nicely for APIs, prototypes, or internal tools.

Hands-On: Setting Up a Modern AI-Friendly Flight Project

Here’s a practical starter for newcomers (and a refresher for everyone else). We’ll use the skeleton and enable AI tools.

1. Prerequisites: PHP 7.4+ (8.3+ recommended for new projects).

2. Create the Project:
   

   composer create-project flightphp/skeleton my-flight-app
   cd my-flight-app

1. Serve Locally:

   composer start

Or: php -S localhost:8000 -t public/. Visit http://localhost:8000.

1. Configure AI:

   php runway ai:init

Follow prompts for your provider, API key, and model.

1. Generate Project Instructions:

   php runway ai:generate-instructions

Answer questions about your app (e.g., SQLite DB, Latte templating, REST focus). This populates files for Copilot, Cursor, etc.

1. Add the MCP Server (for even better context): In your AI tool (e.g., Claude or Cursor), add the MCP URL: https://mcp.flightphp.com/mcp.

From here, prompt confidently: “Add a grouped /api/tasks route with auth middleware using PDO SQLite and return JSON.” The AI now has the right context.

When to Choose Flight (and When Not To)

Choose Flight for:

• APIs and microservices.
• Learning or teaching web fundamentals.
• Projects where you want full control and minimal overhead.
• AI-augmented workflows that reward clarity.

Consider alternatives like the Slim AI skeleton (or Laravel for complex enterprise needs) if your team or AI setup prefers different conventions. The PHP ecosystem is rich—use what keeps you shipping delightfully.

The Verdict: Not Dead, Just Different

Flight PHP is alive and well in 2026. Its recent AI investments, tiny core, and focus on simplicity give it a bright spot in an ecosystem dominated by heavier options. The PHP 7.4 usage spike reminds us that not every project needs (or wants) the latest bells and whistles—many just need something reliable and fast.

If you’re on the fence, try the skeleton. Experiment with Runway and the MCP server. You might find, like many of us have over the years, that “simple” is surprisingly powerful.

What LLM setup are you using with Flight (or Slim)? Drop your experiences in the comments—I read them all.