MCP Server

Overview

The LaraDumps MCP (Model Context Protocol) Server is built into the LaraDumps Desktop Application and provides AI-powered debugging capabilities by exposing debugging data through the Model Context Protocol. This allows Cursor, OpenCode, and other MCP-compatible AI models to access debugging information about your Laravel application in real-time.

The MCP Server runs in HTTP/SSE mode through LaraDumps Settings, making it easy to connect with these clients.

1. Getting Started

Follow these steps to set up the LaraDumps MCP Server with your AI editor.

2. Enable the MCP Server

1. Open LaraDumps Desktop Application
2. Go to Settings → MCP Server tab
3. Toggle Enable MCP Server
4. Note the port number (default: 3000)

3. Limit Payload Objects

Controls the maximum number of items returned by each tool call (logs, queries, dumps, etc.).

Configure in: LaraDumps → Settings → MCP Server → Limit Payload objects
Default: 300
Purpose: Prevents overwhelming AI models with too much data at once.

Configure Your Client

Cursor

1. Open Cursor Settings (cmd + ,)
2. Go to Features > MCP
3. Click Add New MCP Server
4. This will open the mcp.json file. Add the following configuration:

{
  "mcpServers": {
    "laradumps": {
      "url": "http://127.0.0.1:3000/sse",
      "type": "sse"
    }
  }
}

(Replace 3000 with your configured port if different)

OpenCode

Add the following configuration to your mcpServers settings:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "LaraDumps": {
      "type": "remote",
      "url": "http://127.0.0.1:3000/sse"
    }
  }
}

Verify Connection

1. Restart your editor (Cursor or OpenCode) to apply changes.
2. Open a chat with the AI.
3. You should see LaraDumps tools available (e.g., get_logs, get_queries).

Configuration Details

Port Configuration

The MCP Server port is configured through LaraDumps → Settings → MCP Server. Default: 3000.

API Base URL

The MCP server communicates internally with LaraDumps via http://0.0.0.0:9191/api/mcp.

Available Tools

Retrieval Tools

Tool Name	Description
get_jobs	Retrieve all captured background jobs from your Laravel application.
get_brains	Get collected "Brains" data (insights and debugging metadata).
get_dumps	Get all captured variable dumps.
get_project_info	Retrieve metadata including PHP version, Laravel version, and environment.
get_mails	Get all captured emails sent by your application.
get_livewire_components	Get all registered Livewire components in your application.
get_queries	Retrieve captured SQL queries. Supports limit parameter.
get_logs	Get a summary of all captured application logs.
get_log_details	Retrieve full details of a specific log entry by log_id.
summarize_logs	Get a summary of the last 50 application logs.

Action Tools

Tool Name	Description
confetti	Trigger confetti animation in the LaraDumps desktop application.
clear_jobs	Clear all captured background jobs from memory.
clear_mails	Clear all captured emails from the dashboard.
clear_logs	Clear all captured application logs.
clear_dumps	Clear all captured variable dumps.
toggle_env	Enable, disable or toggle a environment/watcher (e.g. queries, logs, cache).

Available Prompts

Prompts are pre-configured AI interactions that gather relevant debugging data and present it to the AI model for analysis.

analyze_last_exception

Analyzes the most recent error or exception in your application logs.

Behavior:

• Fetches all logs
• Filters for error-level logs (error, critical, emergency, alert)
• Presents the most recent error to the AI for analysis
• Prompts the AI to suggest fixes

Example Use Case: Quickly understand what caused an error and get AI-powered solutions.

optimize_latest_query

Analyzes the most recent SQL query for performance optimization.

Behavior:

• Fetches all captured queries
• Extracts the latest query
• Provides the query to the AI for performance analysis
• Prompts the AI to suggest optimization strategies

Example Use Case: Identify N+1 queries, missing indexes, or inefficient SQL patterns.

Common Commands

1. Real-time Code Analysis

Use Cursor or OpenCode with MCP to analyze your latest queries and logs as you're developing:

"Analyze my last query and tell me if there are any N+1 problems"
→ Uses `get_queries` and `optimize_latest_query` prompt

2. Error Debugging

Quickly understand what went wrong:

"What caused the last error in my application?"
→ Uses `get_logs` and `analyze_last_exception` prompt

3. Performance Optimization

Get AI-powered suggestions for improving your application:

"Are my background jobs running efficiently?"
→ Uses `get_jobs` and `get_project_info` tools

4. Monitoring Control

Enable/disable specific monitoring features through AI:

"Stop monitoring queries for now"
→ Uses `toggle_env` tool with queries

5. Project Understanding

Get comprehensive project information:

"What's running in my Laravel application?"
→ Uses `get_project_info`, `get_livewire_components`, `get_mails`

Environment Variables

The MCP server respects the following environment configuration:

Variable	Purpose	Default
API_BASE	LaraDumps API endpoint	http://0.0.0.0:9191/api/mcp

Performance Considerations

• Query Limits: The get_queries tool limits to 50 results by default to avoid overwhelming data transfer
• Real-time Updates: Each tool call fetches fresh data from LaraDumps
• Response Sizes: Large responses (many logs/queries) may take longer to process
• Network Latency: Minimal overhead for local HTTP connections

Troubleshooting

MCP Server Not Connecting

Problem: The client doesn't see the LaraDumps MCP Server.

Solution:

1. Verify the MCP Server is enabled in LaraDumps → Settings → MCP Server
2. Check the port number matches your client config
3. Ensure the URL in your config is http://127.0.0.1:PORT/sse
4. Restart your editor (Cursor/OpenCode)
5. Try starting a new conversation

Connection Refused Error

Problem: "Connection refused" when trying to connect.

Solution:

1. Make sure the LaraDumps desktop app is running
2. Verify MCP Server is enabled in Settings
3. Confirm the port number is correct
4. Test locally: curl http://127.0.0.1:3000/sse (replace 3000 with your port)

No Data Returned

Problem: Tools run but return empty results.

Solution:

1. Verify your Laravel app is sending data to LaraDumps
2. Check that you have enabled monitoring for the data types you want
3. Look in the LaraDumps desktop app to see if data is being captured

Tools Aren't Responding

Problem: The client calls tools but gets timeout or no response.

Solution:

1. Ensure the MCP Server is enabled in LaraDumps Settings
2. Verify your Laravel app is running and sending data
3. Wait 10 seconds after restarting your editor
4. Try starting a new conversation