mcp-server-graylog

Model Context Protocol (MCP) server for Graylog log searching. Search logs by absolute/relative timestamps, filter by streams, and debug production issues directly from Claude Desktop.

Built for production debugging - Search Graylog logs using exact timestamps, filter by application streams, and get actionable insights for troubleshooting production issues.

Features

• ✅ Absolute timestamp search - Debug specific errors with exact time ranges
• ✅ Relative timestamp search - Search recent logs (last N seconds)
• ✅ Distributed tracing - Follow a trace_id across all services
• ✅ Surrounding-log context - See what happened ±N seconds around an error
• ✅ Composite incident analysis - One tool call fans out to trace + context + baseline
• ✅ Field aggregation - Group counts by service/level/pod/lead_id with bandwidth-efficient projection
• ✅ Stream discovery - List all available streams/applications
• ✅ System health check - Verify Graylog connectivity
• ✅ Comprehensive validation - ISO 8601 timestamps, query syntax, stream IDs
• ✅ Clear error messages - Actionable errors for auth, network, and API issues
• ✅ Timeout handling - 30-second timeouts prevent hanging
• ✅ Production-ready - 54 tests, 9.2/10 code quality score

Table of Contents

• Installation
• Configuration
• Available Tools
• Skills & agents
• Query Examples
• Troubleshooting
• Development
• Contributing
• License

Installation

Option 1: Use with npx (Recommended)

# No installation needed - use directly with npx
npx mcp-server-graylog

Option 2: Global Installation

npm install -g mcp-server-graylog

Option 3: Local Installation

# Clone the repository
git clone https://github.com/Pranavj17/mcp-server-graylog.git
cd mcp-server-graylog

# Install dependencies
npm install

Configuration

Claude Desktop Setup

Add to your Claude Desktop config file:

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

Using npx (Recommended)

{
  "mcpServers": {
    "graylog": {
      "command": "npx",
      "args": ["-y", "mcp-server-graylog"],
      "env": {
        "BASE_URL": "https://graylog.example.com",
        "API_TOKEN": "your_api_token_here"
      }
    }
  }
}

Using Local Installation

{
  "mcpServers": {
    "graylog": {
      "command": "node",
      "args": ["/path/to/mcp-server-graylog/src/index.js"],
      "env": {
        "BASE_URL": "https://graylog.example.com",
        "API_TOKEN": "your_api_token_here"
      }
    }
  }
}

Environment Variables

Getting Your Graylog API Token

1. Log in to Graylog web interface
2. Go to System → Users
3. Select your user
4. Click Edit tokens
5. Create a new token with read permissions
6. Copy the token value

Available Tools

1. search_logs_absolute

Search logs using absolute timestamps (from/to). Perfect for debugging errors with specific timestamps from monitoring tools or error tracking systems.

Parameters:

• query (required): Search query using Elasticsearch syntax
• from (required): Start timestamp in ISO 8601 format
• to (required): End timestamp in ISO 8601 format
• streamId (optional): Stream ID to filter results
• limit (optional): Maximum results (default: 50, max: 1000)

Example:

{
  "query": "\"/api/v1/registrations\" AND \"PUT\"",
  "from": "2025-10-23T10:00:00.000Z",
  "to": "2025-10-23T11:00:00.000Z",
  "streamId": "646221a5bd29672a6f0246d8",
  "limit": 100
}

2. search_logs_relative

Search logs using relative time range (e.g., last 15 minutes). Useful for recent log analysis.

Parameters:

• query (required): Search query using Elasticsearch syntax
• rangeSeconds (optional): Time range in seconds (default: 900 = 15 minutes, max: 86400 = 24 hours)
• streamId (optional): Stream ID to filter results
• limit (optional): Maximum results (default: 50, max: 1000)

Example:

{
  "query": "level:ERROR",
  "rangeSeconds": 3600,
  "limit": 100
}

3. trace_request

Trace a request across ALL services using a trace_id. Fetches logs from every stream, groups by service/pod, and sorts each service's messages chronologically. Essential for distributed debugging in microservice architectures.

Parameters:

• traceId (required): The trace ID to follow (e.g., abbb27610a7fd76be8fb5af17edbe00d)
• from (required): Start timestamp in ISO 8601 format (search window)
• to (required): End timestamp in ISO 8601 format (search window)
• limit (optional): Maximum results (default: 200, max: 1000)

Example:

{
  "traceId": "abbb27610a7fd76be8fb5af17edbe00d",
  "from": "2026-05-13T15:38:00.000Z",
  "to":   "2026-05-13T15:48:00.000Z"
}

4. get_surrounding_logs

Return logs within ±N seconds of a timestamp, optionally filtered by source/pod/stream. Reveals what happened immediately before and after an error.

Parameters:

• timestamp (required): Center timestamp in ISO 8601 format
• source (optional): Source hostname or pod to filter by
• streamId (optional): Stream ID filter
• windowSeconds (optional): Window on each side (default: 5, max: 300)
• limit (optional): Maximum results (default: 100)

Example:

{
  "timestamp": "2026-05-13T15:43:27.844Z",
  "source": "argus-production-f747f5d4d-x9hpp",
  "windowSeconds": 10
}

5. analyze_incident

Composite tool. One call fans out to three searches and returns an aggregated incident report — saves 2-3 LLM orchestration rounds when investigating a specific trace.

Internally executes:

1. The full trace hop chain (trace_id:X)
2. Pod-scoped surrounding logs around the first ERROR/CRITICAL/FATAL hop (filters by pod: to avoid multi-tenant noise on shared hosts)
3. A trailing-hour error baseline for the anchor service

Parameters:

• traceId (required): The trace ID to investigate
• from (required): Start timestamp in ISO 8601 format
• to (required): End timestamp in ISO 8601 format
• window (optional): Surrounding-logs window in seconds (default: 10, max: 300)
• baselineSeconds (optional): Trailing window for the baseline lookup (default: 3600, max: 86400)

Example:

{
  "traceId": "abbb27610a7fd76be8fb5af17edbe00d",
  "from": "2026-05-13T15:38:00.000Z",
  "to":   "2026-05-13T15:48:00.000Z",
  "window": 10,
  "baselineSeconds": 3600
}

Returns (abridged):

{
  "trace_id": "abbb27610a7fd76be8fb5af17edbe00d",
  "found": true,
  "steps_executed": 4,
  "summary": {
    "hops": 4,
    "services_involved": ["argus"],
    "errors_in_trace": 1,
    "anchor_service": "argus",
    "anchor_pod": "argus-production-f747f5d4d-x9hpp",
    "first_error": { "timestamp": "...", "service": "argus", "message": "nil fund_id ...", "lead_id": "..." },
    "request": { "http_path": "/api/v2/user/graph", "http_method": "POST", "http_status": 200, "duration_ms": 67 },
    "baseline_errors_in_service": 16,
    "baseline_window_seconds": 3600
  },
  "trace_hops": [...],
  "surrounding_logs": [...]
}

6. aggregate_logs

Count log entries grouped by a field — Graylog's most-used operation, made one-call. Issues a single search with fields=<group_field> projected (so only the column you want is downloaded) and aggregates client-side. Replaces Graylog 5.x's removed legacy terms-aggregation endpoint.

Parameters:

• query (required): Filter (Elasticsearch syntax). Use * for all entries.
• field (required): Field to group by. Common: service, logger_level, pod, lead_id, http_status, container_name.
• from+to OR rangeSeconds (required, mutually exclusive): time window
• size (optional): Top N to return (default 25, max 100). Rest summed into other.
• fetchLimit (optional): Max messages to aggregate (default 5000, max 10000). When matched exceeds this, truncated: true is flagged.
• streamId (optional)

Example:

{
  "query": "logger_level:error",
  "field": "service",
  "rangeSeconds": 1800,
  "size": 10
}

Returns:

{
  "field": "service",
  "query": "logger_level:error",
  "time_range": "Last 1800 seconds",
  "total_matched": 30,
  "messages_aggregated": 30,
  "truncated": false,
  "unique_groups": 5,
  "top": { "milkyway": 8, "argus": 4, "telex": 4, "advisory": 3, "auth": 1 },
  "other": 0,
  "missing": 10,
  "api_calls": 1
}

The missing count is messages that matched the query but had no value for the group-by field — useful signal for log-hygiene issues.

7. list_streams

List all available Graylog streams (applications). Use this to discover stream IDs for filtering.

Parameters: None

Returns:

{
  "total": 3,
  "streams": [
    {
      "id": "646221a5bd29672a6f0246d8",
      "title": "application-api",
      "description": "API application logs",
      "disabled": false
    }
  ]
}

8. get_system_info

Get Graylog system information and health status. Verify connectivity and check server version.

Parameters: None

Returns:

{
  "version": "5.1.0",
  "codename": "graylog",
  "cluster_id": "abc123",
  "is_processing": true,
  "timezone": "UTC"
}

Skills & agents (v2.3.0+)

When installed as a Claude Code plugin, this package ships playbooks that teach Claude when and how to use the MCP tools above.

Skills

Agent

Skills auto-load when the plugin is installed. The agent is dispatchable via Claude Code's subagent mechanism with subagent_type: "graylog-trace-analyzer".

Query Examples

Search for Errors

level:ERROR

Search for Specific Endpoint

"/api/v1/registrations" AND "PUT"

Search for HTTP Status Codes

status:500
status:>=400

Search for User Actions

user_id:12345 AND action:login

Search for Slow Requests

duration_ms:>1000

Search for Exceptions

exception:NullPointerException

Combine Multiple Conditions

level:ERROR AND source:nexus AND message:*timeout*

Search with Wildcards

message:*connection refused*

Search by Field Existence

_exists_:error_code

Common Use Cases

1. Debug Production Error

When you get an error with a timestamp from your monitoring system:

1. Copy error timestamp from your monitoring tool
2. Use search_logs_absolute with ±5 minute window
3. Filter by application stream
4. Find root cause in logs

2. Monitor Recent Deployments

After deploying:

1. Use search_logs_relative with last 15 minutes
2. Search for level:ERROR
3. Verify no new errors introduced

3. Investigate API Failures

When an API endpoint fails:

1. Search for endpoint path: "/api/v1/endpoint"
2. Filter by status codes: status:>=400
3. Check error patterns

Error Messages

The server provides clear, actionable error messages:

Troubleshooting

Server Won't Start

Check environment variables:

# Verify BASE_URL and API_TOKEN are set in Claude Desktop config
# Check Claude Desktop logs:
# macOS: ~/Library/Logs/Claude/mcp*.log
# Windows: %APPDATA%\Claude\logs\mcp*.log

Verify Graylog accessibility:

curl -u "YOUR_API_TOKEN:token" https://graylog.example.com/api/system

Authentication Errors

• Verify API token has read permissions in Graylog
• Token format: Use token value as username, "token" as password
• Check token hasn't expired

No Results Returned

• Verify stream ID is correct using list_streams tool
• Check timestamp range includes data
• Try simplifying query to * to see if any data exists
• Verify stream is not disabled

Integration Tests Failing

# Set environment variables for integration tests
export INTEGRATION_TESTS=true
export BASE_URL=https://graylog.example.com
export API_TOKEN=your_token_here

# Run integration tests
npm run test:integration

Development

Prerequisites

• Node.js >= 18.0.0
• npm >= 8.0.0
• Access to a Graylog instance (for integration tests)

Development Workflow

# Install dependencies
npm install

# Run in development mode (auto-reload)
npm run dev

# Run tests
npm test

# Run tests in watch mode
npm run test:watch

# Run only unit tests
npm run test:unit

# Run integration tests (requires Graylog instance)
INTEGRATION_TESTS=true BASE_URL=https://graylog.example.com API_TOKEN=xxx npm run test:integration

# Check syntax
npm run lint

Project Structure

mcp-server-graylog/
├── src/
│   └── index.js           # Main server implementation (429 lines)
├── test/
│   ├── helpers.test.js    # Helper function tests (14 tests)
│   ├── validation.test.js # Input validation tests (24 tests)
│   ├── mcp-protocol.test.js # MCP protocol tests (16 tests)
│   └── integration.test.js  # Integration tests (7 tests)
├── example-config.json    # Claude Desktop config example
├── CONTRIBUTING.md        # Contributing guidelines
├── CHANGELOG.md          # Version history
└── package.json         # npm configuration

Running Tests

# Run all tests (54 tests)
npm test

# Expected output:
# tests 54
# pass 54
# fail 0

Architecture

Simple, focused architecture in a single file (429 lines):

• Configuration & Validation - Environment variable checking
• Helper Functions - ISO 8601 validation, error formatting
• MCP Server Setup - Standard MCP protocol implementation
• Tool Definitions - 4 tools with clear schemas
• Tool Implementations - Clean, validated functions
• Server Startup - Validation then connection

Design Principles:

• ✓ Simple and maintainable
• ✓ One file, easy to understand
• ✓ Clear separation of concerns
• ✓ Comprehensive error handling
• ✓ Input validation at boundaries
• ✓ Consistent response format

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

Quick Start:

1. Fork the repository
2. Create a feature branch
3. Add tests for your changes
4. Ensure all tests pass (npm test)
5. Submit a pull request

Changelog

See CHANGELOG.md for version history and release notes.

Security

• Environment variables for sensitive data (never hardcoded)
• Basic authentication properly implemented
• Input validation prevents injection attacks
• Timeout prevents hanging requests
• Error messages don't leak sensitive information

To report security vulnerabilities, please create a private security advisory on GitHub.

License

MIT License - see LICENSE file for details.

Links

• npm package
• GitHub repository
• Issue tracker
• Model Context Protocol
• Graylog Documentation

Acknowledgments

• Built with @modelcontextprotocol/sdk
• Inspired by the MCP community
• Thanks to all contributors!

Made with ❤️ for the Claude Desktop community

For questions or support, please open an issue on GitHub