- Is killerwolf/mcp-server-bundle compatible with Laravel?
- No. The package is a Symfony bundle, not a Laravel package, and it depends on the Symfony full-stack runtime, including Kernel, the DI container, EventDispatcher, the HttpKernel request/response lifecycle, and the file-based Profiler storage layer. Laravel does not provide those contracts, so the bundle will not register against Laravel's service provider lifecycle. Installing it in a Laravel app would require either a Symfony micro-kernel or a custom compatibility shim, both more work than writing a purpose-built MCP server.
- Can I install killerwolf/mcp-server-bundle with Composer in a Laravel project?
- Composer-level installation will pull Symfony components and a bundle that will not register against Laravel's service provider lifecycle. Downloading the package is possible, but integration is not supported without a Symfony micro-kernel or a compatibility shim. The technical assessment recommends against depending on it as an installable dependency.
- What does killerwolf/mcp-server-bundle actually do?
- It exposes Symfony profiler data such as requests, queries, and timings through the Model Context Protocol so AI clients can discover and call tools and resources over standardized MCP endpoints. It is a read-only integration that surfaces observability payloads rather than modifying your application. Its value is tied to the Symfony Profiler storage layer, which has no direct equivalent data source in Laravel.
- Why is this package a poor fit for Laravel's observability stack?
- Laravel's observability surface is built around Telescope, Debugbar, Pulse, and the query, event, and job listeners those provide. The bundle binds to Symfony Profiler storage, so there is no direct data source to expose in a Laravel stack. The mismatch is at the foundation layer, not merely the adapter layer.
- Is killerwolf/mcp-server-bundle still maintained?
- No. Version 0.2.1 is explicitly the final release, and the release notes state that no further updates are planned. The sanctioned successor, Symfony AI Mate, is also Symfony-only, with no Laravel successor indicated. The package should be treated as deprecated.
- Does the package work with Laravel Telescope or Debugbar?
- No, not natively. It reads from the Symfony Profiler token store, not from Telescope or Debugbar storage. Making it work with Laravel data would require a shim that fakes the Profiler storage contract against Telescope data, which the assessment describes as more work than building a purpose-built MCP server.
- What are the security risks of exposing profiler data over MCP?
- Profiler payloads can include request data, SQL queries, and timings, so exposing them to AI clients is sensitive. The technical assessment notes that the transport and auth model are unverified from the package surface: stdio runs as a local subprocess per developer, while HTTP/SSE would be a shared service with a categorically different risk profile. Read-only scope limits blast radius because the tool does not write to the application, but a shared endpoint exposing request payloads and SQL is still high risk.
- Which PHP versions does killerwolf/mcp-server-bundle support?
- PHP 8.4 compatibility was patched in with an explicit nullable type fix in MCPDataCollector::collect(), and CI was restored for PHP 8.1 through 8.4. That means it will not break on a PHP 8.4 upgrade if you already run it. It does not extend the maintenance horizon, since the package is final at 0.2.1.
- How do I remove killerwolf/mcp-server-bundle from an existing install?
- The documented exit path is three steps: run composer remove for the package, delete the bundle line from config/bundles.php, and remove the mcp:server:run entry from your MCP client config. No data migration is needed because the profiler store is disposable by nature. The assessment describes this as a well-handled deprecation that lowers the cost of an abandoned experimental install.
- What should I use instead for MCP in Laravel, and can this package serve as a reference?
- There is no Laravel successor to this bundle; Symfony AI Mate is Symfony-only. The reusable assets are the tool schema design, meaning which profiler facts to expose and at what granularity, plus the Agent Skills packaging model that Mate introduces, which bundles domain knowledge as discoverable, progressive-disclosure skills rather than a flat tool list. Both are transferable to a Laravel-native MCP server, but the code itself is not worth depending on.