logpare

MCP Integration

Integrate logpare with AI coding assistants via the Model Context Protocol

Learn how to integrate logpare with AI coding assistants via the Model Context Protocol (MCP).

Overview

logpare provides an MCP server that exposes log compression capabilities as tools for AI assistants like Claude Desktop. This enables AI agents to:

  • Compress large log files before analysis
  • Extract patterns from application logs
  • Estimate compression ratios
  • Process logs incrementally

Installation

Global Installation

Install the MCP server globally:

npm install -g @logpare/mcp

Local Installation

Or install in a project:

npm install @logpare/mcp

Claude Desktop Configuration

Add the logpare MCP server to Claude Desktop's configuration.

macOS

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "logpare": {
      "command": "npx",
      "args": ["@logpare/mcp"]
    }
  }
}

Windows

Edit %APPDATA%\Claude\claude_desktop_config.json:

{
  "mcpServers": {
    "logpare": {
      "command": "npx",
      "args": ["@logpare/mcp"]
    }
  }
}

Linux

Edit ~/.config/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "logpare": {
      "command": "npx",
      "args": ["@logpare/mcp"]
    }
  }
}

Verify Installation

Restart Claude Desktop and check that the logpare tools are available:

  1. Open Claude Desktop
  2. Start a new conversation
  3. Type /tools to see available tools
  4. Look for compress_logs, analyze_log_patterns, etc.

Available Tools

compress_logs

Compress an array of log lines with full options.

Parameters:

{
  lines: string[];           // Log lines to compress
  format?: 'summary' | 'detailed' | 'json';
  depth?: number;            // Default: 4
  simThreshold?: number;     // Default: 0.4
  maxTemplates?: number;     // Default: 50
}

Example Usage:

User: Can you compress these logs for me?
[paste logs here]

Claude: I'll use the logpare MCP tool to compress those logs.
[Claude calls compress_logs tool]

Here's the compressed output showing 15 unique templates...

analyze_log_patterns

Extract patterns without full compression (faster for quick analysis).

Parameters:

{
  lines: string[];
  maxPatterns?: number;      // Default: 20
}

estimate_compression

Quick estimate of compression ratio without full processing.

Parameters:

{
  lines: string[];
  sampleSize?: number;       // Default: 1000
}

Usage Examples

Basic Log Analysis

User prompt:

Here are my application logs from the last hour. Can you identify the main issues?

[paste logs]

Claude's approach:

  1. Calls compress_logs with default settings
  2. Analyzes the compressed templates
  3. Identifies error patterns and frequencies
  4. Provides actionable insights

Comparing Log Files

User prompt:

Compare these two log files and tell me what changed:

File 1: [logs from yesterday]
File 2: [logs from today]

Claude's approach:

  1. Calls compress_logs for each file
  2. Compares template sets
  3. Identifies new patterns, removed patterns, frequency changes

Performance Investigation

User prompt:

These logs are from a slow API endpoint. What's causing the delays?

[paste logs]

Claude's approach:

  1. Calls compress_logs with format: 'detailed'
  2. Examines durationSamples in templates
  3. Identifies patterns with high latency
  4. Suggests optimizations

Configuration Options

Custom MCP Server Settings

You can configure the MCP server with environment variables or a config file.

Create .logpare-mcp.json in your home directory:

{
  "defaultFormat": "detailed",
  "defaultDepth": 5,
  "defaultSimThreshold": 0.4,
  "maxLinesPerRequest": 100000,
  "enableProgressReporting": true
}

Environment Variables

# Set default output format
export LOGPARE_MCP_FORMAT=detailed
 
# Set default depth
export LOGPARE_MCP_DEPTH=5
 
# Set max lines per request
export LOGPARE_MCP_MAX_LINES=100000

Update your Claude Desktop config:

{
  "mcpServers": {
    "logpare": {
      "command": "npx",
      "args": ["@logpare/mcp"],
      "env": {
        "LOGPARE_MCP_FORMAT": "detailed",
        "LOGPARE_MCP_DEPTH": "5"
      }
    }
  }
}

Best Practices

1. Start with Compression Estimates

For very large files, estimate first:

User: I have 500,000 log lines. Should I compress them all?

Claude: Let me estimate first...
[calls estimate_compression with sample]

Based on the sample, compression will be effective. I'll proceed with full compression.

2. Use Appropriate Formats

  • summary: Quick overview of main patterns
  • detailed: Deep analysis with metadata
  • json: When you need programmatic access

3. Tune Parameters for Log Type

Ask Claude to adjust based on your logs:

User: These are Kubernetes pod logs, very noisy

Claude: I'll use depth=5 and simThreshold=0.3 for better grouping
[adjusts parameters accordingly]

4. Leverage Metadata

Extract specific information from templates:

User: What URLs are appearing in these error logs?

Claude: [uses compress_logs with format='json']
Looking at the urlSamples in error templates:
- api.example.com (450 errors)
- cdn.example.com (23 errors)

Troubleshooting

MCP Server Not Appearing

  1. Check Claude Desktop config syntax (valid JSON)
  2. Restart Claude Desktop completely
  3. Check MCP server logs:

macOS/Linux:

tail -f ~/Library/Logs/Claude/mcp*.log

Windows:

Get-Content "$env:APPDATA\Claude\Logs\mcp*.log" -Wait

Tools Not Working

  1. Verify @logpare/mcp is installed:
npm list -g @logpare/mcp
  1. Test the server directly:
npx @logpare/mcp --test
  1. Check for error messages in Claude Desktop

Performance Issues

If compression is slow:

  1. Reduce maxTemplates in requests
  2. Use estimate_compression first for large files
  3. Process in smaller batches
  4. Increase simThreshold to create fewer templates

Security Considerations

Sensitive Data

Be cautious when sharing logs with AI assistants:

  1. PII Masking: Ensure sensitive data is masked
  2. Custom Patterns: Add patterns for proprietary IDs
  3. Local Processing: MCP runs locally - logs stay on your machine
  4. Review Output: Check compressed output before sharing

Example: Masking Sensitive Data

// In your preprocessing strategy
const sensitiveStrategy = defineStrategy({
  preprocess(line: string): string {
    return line
      // Mask emails
      .replace(/\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b/g, '<EMAIL>')
      // Mask phone numbers
      .replace(/\b\d{3}[-.]?\d{3}[-.]?\d{4}\b/g, '<PHONE>')
      // Mask SSN
      .replace(/\b\d{3}-\d{2}-\d{4}\b/g, '<SSN>')
      // Mask credit cards
      .replace(/\b\d{4}[\s-]?\d{4}[\s-]?\d{4}[\s-]?\d{4}\b/g, '<CARD>');
  }
});

See Also

Previous

Custom Preprocessing

Next

CLI Reference

On this page