Laravel Mcp Log

MCP tool for Laravel log analysing with AI.

This package uses the core log reading functionality provided by the Laravel Log Reader, so for the details about how the log reading works, please check that repository.

Installation

You can install the package (that you created with this template) via composer:

composer require laplace-demon-ai/laravel-mcp-log

You can publish the configuration file with:

php artisan vendor:publish --tag="laravel-mcp-log"

Configuration

You can enable/disable the MCP Log Reader as:

'enabled' => env('MCP_LOG_READER_ENABLED', true),

The package supports two log storage drivers: file and db for now.

Database Logs

If your logs are stored in database (e.g. log_entries table), in the config file (laravel-mcp-log.php) set the driver to db as:

'driver' => env('LOG_READER_DRIVER', LogDriverType::DB->value),

And set the other database related configurations as:

'table' => env('LOG_DB_TABLE_NAME', 'log_entries'),
'limit' => env('LOG_READER_DB_QUERY_LIMIT', 10000), // max number of records to fetch in queries

// Column mapping: maps DB columns to LogData properties
'columns' => [
    LogTableColumnType::ID->value => 'id',
    LogTableColumnType::LEVEL->value => 'level', // e.g. 'ERROR', 'INFO'
    LogTableColumnType::MESSAGE->value => 'message', // main log message
    LogTableColumnType::TIMESTAMP->value => 'created_at', // time of the log entry (e.g. 'created_at' or 'logged_at')
    LogTableColumnType::CHANNEL->value => 'channel', // e.g. 'production', 'local'
    LogTableColumnType::CONTEXT->value => 'context', // additional context info, often JSON e.g. '{"action":"UserLogin"}'
    LogTableColumnType::EXTRA->value => 'extra', // any extra data, often JSON e.g. '{"ip":172.0.0.1, "session_id":"abc", "user_id":123}'
],

'searchable_columns' => [
    ['name' => LogTableColumnType::MESSAGE->value, 'type' => ColumnType::TEXT->value],
    ['name' => LogTableColumnType::CONTEXT->value, 'type' => ColumnType::JSON->value],
    ['name' => LogTableColumnType::EXTRA->value, 'type' => ColumnType::JSON->value],
],

Basically we do a mapping between the database columns of your log table and the package, so that it can read the logs properly.

• The searchable_columns defines which columns are searchable and their types. It is used for the search queries. And we require text and json column types so that we choose the right search strategy (default is text).
• The limit defines the maximum number of records to fetch in queries to avoid performance issues.

File Logs

If your logs are stored in files (laravel.log), in the config file (laravel-mcp-log.php) set the driver to file as:

'driver' => env('LOG_READER_DRIVER', LogDriverType::FILE->value),

And set the log file path as:

'path' => env('LOG_FILE_PATH', storage_path('logs/laravel.log')),
'limit' => env('LOG_READER_FILE_QUERY_LIMIT', 10000), // max number of lines to read

• The path defines the full path to your Laravel log file.
• The limit defines the maximum number of lines to read in queries to avoid performance issues

Usage

This package automatically registers MCP Server (LogReaderServer) and MCP tool (LogReaderTool) in LaravelMcpLogServiceProvider, so you don't need to register it manually within the ai.php in your Laravel application.

[!WARNING] You only need to install official Laravel MCP package so that MCP can recognize this server.

We defined the log reader endpoint as following in the LaravelMcpLogServiceProvider:

Mcp::web('mcp/log-reader', LogReaderServer::class);

It means that you can access the log reader MCP tool in your MCP dashboard at your-app-url/mcp/log-reader URL.

Following is the video for a quick demo of how it works:

• This is how you can use localtunnel to expose your local Laravel application to the internet for MCP access:

Localtunnel and Chatgpt

• And this is a demo of how the MCP Log Reader works with sample prompts:

Demo

TODO (this is for Laravel Log Reader )

• Add a log_insights migration/table which will be a normalized, summarized, and searchable table.
• It unifies different log mechanisms into a single canonical format, enabling faster lookups over large data.
• A background task should sync new log data periodically, basically everyday it summarizes the previous day's logs and inserts them into log_insights.
• Be aware that summarization may lose some details (e.g., exact errors or stack traces).
• Add support for cloud log readers (AWS CloudWatch, Azure Monitor, Google Cloud Logging).
• Add streaming responses for overall logic.
• Use a cheap/free model to summarize large log files before search/filter (experimental approach).
• Laravel MCP Log can be integrated into coding agents where periodically logs are analysed by this tool, and issues are caught and reported to different channels (slack etc). And maybe even PRs can be generated by using the logs.

Contributing

Your contributions are welcome! If you'd like to improve this project, simply create a pull request with your changes. Your efforts help enhance its functionality and documentation.

If you find this project useful, please consider ⭐ it to show your support!

Authors

This project is created and maintained by Moe Mizrak.

License

Laravel Package Template is an open-sourced software licensed under the MIT license.