Log Analyzer MCP Server

A Model Context Protocol (MCP) server that connects to the Log Collector API to query and analyze logs from multiple remote Unix hosts.

Features

• MCP server implementation for Claude Desktop and other MCP clients

• Query logs from multiple hosts via REST API

• Advanced log analysis (errors, warnings, patterns)

• Search across multiple hosts and processes

• Statistical analysis of log content

• Summary generation

Related MCP server: Log Analyzer MCP

Installation

1. Create a virtual environment:

cd log-analyzer-mcp
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Configure environment variables:

cp .env.example .env
# Edit .env with your Log Collector API URL

Configuration

Environment Variables (.env)

LOG_API_BASE_URL=http://localhost:8000
LOG_API_TIMEOUT=30

Running the MCP Server

Standalone Mode

python -m log_analyzer_mcp.server

As an MCP Server (for Claude Desktop)

Add to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "log-analyzer": {
      "command": "python",
      "args": ["-m", "log_analyzer_mcp.server"],
      "env": {
        "LOG_API_BASE_URL": "http://localhost:8000",
        "LOG_API_TIMEOUT": "30"
      }
    }
  }
}

Or if installed as a package:

{
  "mcpServers": {
    "log-analyzer": {
      "command": "log-analyzer-mcp"
    }
  }
}

Available MCP Tools

1. list_hosts

Lists all configured hosts and their processes.

Parameters: None

Usage in Claude:

List all available hosts

2. get_logs

Get logs from a specific host with optional process filtering.

Parameters:

• host_name (required): Name of the host

• process (optional): Filter by process name

• lines (optional): Number of lines to retrieve (default: 100)

• tail (optional): Get last N lines (default: true)

Usage in Claude:

Get the last 200 lines from app-server-1
Show logs from app-server-1 for the app1 process

3. get_all_logs

Get logs from all configured hosts.

Parameters:

• lines (optional): Number of lines per file (default: 100)

• tail (optional): Get last N lines (default: true)

Usage in Claude:

Get logs from all hosts
Show recent logs from all servers

4. search_logs

Search for a pattern across logs.

Parameters:

• pattern (required): Search pattern (case-insensitive)

• host (optional): Filter by host name

• process (optional): Filter by process name

• lines (optional): Number of lines to search (default: 1000)

Usage in Claude:

Search for "ERROR" in all logs
Find "connection timeout" in app-server-1
Search for "failed" in the database process

5. analyze_logs

Analyze logs and get detailed statistics.

Parameters:

• host_name (required): Name of the host

• process (optional): Filter by process name

• lines (optional): Number of lines to analyze (default: 500)

Usage in Claude:

Analyze logs from app-server-1
Give me statistics on the app1 process logs

6. find_errors

Find all error messages in logs.

Parameters:

• host_name (required): Name of the host

• process (optional): Filter by process name

• lines (optional): Number of lines to search (default: 1000)

Usage in Claude:

Find all errors in app-server-1
Show me errors from the database process

7. get_log_summary

Get a summary of logs including error/warning counts.

Parameters:

• host_name (required): Name of the host

• process (optional): Filter by process name

• lines (optional): Number of lines to analyze (default: 500)

Usage in Claude:

Summarize logs from app-server-1
Give me a summary of the app1 process

8. search_and_summarize

Search for a pattern across logs and get comprehensive summary in one operation.

Parameters:

• pattern (required): Search pattern (case-insensitive)

• host (optional): Filter by host name

• process (optional): Filter by process name

• lines (optional): Number of lines to search (default: 1000)

Returns:

• Match statistics (total matches, files searched, hosts with matches)

• Error/warning counts in matched results

• Time range analysis (first/last timestamp, duration)

• Process timing breakdown (A-Z duration per process)

• Sample matched lines (preview of first 10 matches)

• Hosts breakdown (match count per host)

Usage in Claude:

Search for "ERROR" and give me a summary
Find "connection timeout" and analyze the results
Search for "OutOfMemory" and show me process timing
Search for "database" in app-server-1 and summarize

Example Output:

Search and Summary for pattern: 'ERROR'
================================================================================

MATCH STATISTICS
--------------------------------------------------------------------------------
Total matches: 45
Files searched: 8
Hosts with matches: app-server-1, app-server-2
Errors in matches: 45
Warnings in matches: 3

HOSTS BREAKDOWN
--------------------------------------------------------------------------------
  app-server-1: 32 matches
  app-server-2: 13 matches

TIME RANGE ANALYSIS
--------------------------------------------------------------------------------
First: 2024-02-13 10:30:00
Last: 2024-02-13 15:45:30
Span: 5h 15m 30s

PROCESS TIMING (A-Z)
--------------------------------------------------------------------------------

app1:
  Start: 2024-02-13 10:30:00
  End: 2024-02-13 15:45:30
  Duration: 5h 15m 30s
  Entries: 32

web:
  Start: 2024-02-13 11:00:00
  End: 2024-02-13 14:20:15
  Duration: 3h 20m 15s
  Entries: 13

SAMPLE MATCHED LINES
--------------------------------------------------------------------------------
1. 2024-02-13 10:30:00 ERROR: Connection timeout
2. 2024-02-13 10:35:15 ERROR: Retry failed
...

Usage Examples with Claude

Example 1: Check for Errors Across All Servers

User: Are there any errors in our application servers?

Claude: I'll check for errors across all servers.
[Uses find_errors and analyze_logs tools]

Based on the log analysis:
- app-server-1: 5 errors found in the last 1000 lines
- app-server-2: 2 errors found
Most common error: "Database connection timeout"

Example 2: Investigate a Specific Issue

User: Search for "OutOfMemory" errors in the app1 process

Claude: [Uses search_logs with pattern="OutOfMemory" and process="app1"]

Found 3 occurrences of OutOfMemory in app1:
- 2026-02-13 10:15:23 - OutOfMemoryError: Java heap space
- 2026-02-13 10:16:45 - OutOfMemoryError: Java heap space
- 2026-02-13 10:18:12 - OutOfMemoryError: Java heap space

Example 3: Get Overview of All Systems

User: Give me an overview of all our systems

Claude: [Uses list_hosts and get_log_summary for each host]

System Overview:
1. app-server-1 (192.168.1.10)
   - Processes: app1, app2
   - Last 500 lines: 12 errors, 45 warnings

2. app-server-2 (192.168.1.11)
   - Processes: service
   - Last 500 lines: 2 errors, 8 warnings

Log Analysis Features

The analyzer automatically detects:

Error Patterns

• ERROR level messages

• Exception stack traces

• FATAL/CRITICAL messages

• "Failed" operations

Warning Patterns

• WARN/WARNING messages

• Deprecated features

• Potential issues

Statistics

• Total line count

• Error/warning counts

• Timestamp extraction

• Common message patterns

• Time range analysis

Architecture

┌─────────────────┐
│  Claude Desktop │
│   (MCP Client)  │
└────────┬────────┘
         │
         │ MCP Protocol
         │
┌────────▼────────────┐
│  Log Analyzer MCP   │
│      Server         │
└────────┬────────────┘
         │
         │ HTTP/REST
         │
┌────────▼────────────┐
│  Log Collector API  │
└────────┬────────────┘
         │
         │ SSH (via Jump Server)
         │
┌────────▼────────────┐
│  Unix Hosts         │
│  (Log Files)        │
└─────────────────────┘

Development

Project Structure

log-analyzer-mcp/
├── log_analyzer_mcp/
│   ├── __init__.py
│   ├── server.py           # MCP server implementation
│   ├── api_client.py       # REST API client
│   └── analyzer.py         # Log analysis utilities
├── pyproject.toml          # Package configuration
├── requirements.txt        # Dependencies
├── .env.example           # Environment template
└── README.md

Adding New Tools

1. Add tool definition in list_tools() function

2. Implement handler in call_tool() function

3. Update documentation

Testing

# Test the API client
python -c "from log_analyzer_mcp.api_client import LogCollectorAPIClient; import asyncio; client = LogCollectorAPIClient('http://localhost:8000'); print(asyncio.run(client.get_hosts()))"

Troubleshooting

MCP Server Not Starting

• Check that the Log Collector API is running

• Verify LOG_API_BASE_URL in .env

• Check Python version (>=3.10 required)

Connection Refused

• Ensure Log Collector API is accessible

• Check firewall settings

• Verify API URL and port

No Results Returned

• Verify hosts are configured in Log Collector API

• Check SSH connectivity in Log Collector API

• Review API logs for errors

Integration with Claude Desktop

Once configured, you can use natural language with Claude:

• "Show me recent errors from production"

• "Analyze logs from app-server-1"

• "Search for database connection issues"

• "What's happening on all servers?"

• "Find OutOfMemory errors in the last hour"

Claude will automatically select and use the appropriate MCP tools to fulfill your requests.

Security Considerations

• The MCP server connects to the REST API (not directly to hosts)

• All SSH security is handled by the Log Collector API

• Use localhost or secure networks for API communication

• Implement authentication on the REST API in production

• Limit log line counts to prevent excessive data transfer

Prerequisites

1. Log Collector API must be running and accessible

2. Hosts must be configured in the API's config/hosts.yaml

3. SSH connectivity must be working (test via API first)

4. Python 3.10 or higher

License

MIT