Datadog MCP Server

A Model Context Protocol (MCP) server for interacting with the Datadog API.

Features

Monitoring: Access monitor data and configurations
Dashboards: Retrieve and view dashboard definitions
Metrics: Query available metrics and their metadata
Events: Search and retrieve events within timeframes
Logs: Search logs with advanced filtering and sorting options
Incidents: Access incident management data
API Integration: Direct integration with Datadog's v1 and v2 APIs
Comprehensive Error Handling: Clear error messages for API and authentication issues
Service-Specific Endpoints: Support for different endpoints for logs and metrics

Prerequisites

Node.js (version 16 or higher)
Datadog account with:
- API key - Found in Organization Settings > API Keys
- Application key - Found in Organization Settings > Application Keys

Application Key Scopes

For improved security, you can scope your Application Key to grant only the minimum permissions required by this MCP server. By default, Application Keys inherit all permissions from the user who created them, but scoped Application Keys allow you to follow the principle of least privilege.

Required Scopes

The following scopes are required for the corresponding features:

Tool(s)	Required Scope	Description
`get-monitors`, `get-monitor`	`monitors_read`	Read access to monitor configurations and states
`get-dashboards`, `get-dashboard`	`dashboards_read`	Read access to dashboard definitions
`get-metrics`, `get-metric-metadata`	`metrics_read`	Read access to metrics list and metadata
`get-events`	`events_read`	Read access to events from the event stream
`search-logs`, `aggregate-logs`	`logs_read_data`	Read access to log data for search and aggregation
`get-incidents`	`incident_read`	Read access to incident management data

Creating a Scoped Application Key

Go to Organization Settings > Application Keys
Click New Key
Enter a name (e.g., "MCP Server - Read Only")
Under Scopes, select only the permissions you need:
- For full functionality: monitors_read, dashboards_read, metrics_read, events_read, logs_read_data, incident_read
- For logs only: logs_read_data
- For monitoring only: monitors_read, dashboards_read, metrics_read
Click Create Key

Note: If you don't specify any scopes when creating an Application Key, it will have full access with all permissions of the creating user. For production use, we recommend always specifying explicit scopes.

Installation

Via npm (recommended)

npm install -g datadog-mcp-server

From Source

Clone this repository
Install dependencies:
```
npm install
```
Build the project:
```
npm run build
```

Configuration

You can configure the Datadog MCP server using either environment variables or command-line arguments.

Environment Variables

Create a .env file with your Datadog credentials:

DD_API_KEY=your_api_key_here
DD_APP_KEY=your_app_key_here
DD_SITE=datadoghq.com
DD_LOGS_SITE=datadoghq.com
DD_METRICS_SITE=datadoghq.com

Note: DD_LOGS_SITE and DD_METRICS_SITE are optional and will default to the value of DD_SITE if not specified.

Command-line Arguments

Basic usage with global site setting:

datadog-mcp-server --apiKey=your_api_key --appKey=your_app_key --site=datadoghq.eu

Advanced usage with service-specific endpoints:

datadog-mcp-server --apiKey=your_api_key --appKey=your_app_key --site=datadoghq.com --logsSite=logs.datadoghq.com --metricsSite=metrics.datadoghq.com

Note: Site arguments don't need https:// - it will be added automatically.

Regional Endpoints

Different Datadog regions have different endpoints:

US (Default): datadoghq.com
EU: datadoghq.eu
US3 (GovCloud): ddog-gov.com
US5: us5.datadoghq.com
AP1: ap1.datadoghq.com

Usage with Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "datadog": {
      "command": "npx",
      "args": [
        "datadog-mcp-server",
        "--apiKey",
        "<YOUR_API_KEY>",
        "--appKey",
        "<YOUR_APP_KEY>",
        "--site",
        "<YOUR_DD_SITE>(e.g us5.datadoghq.com)"
      ]
    }
  }
}

For more advanced configurations with separate endpoints for logs and metrics:

{
  "mcpServers": {
    "datadog": {
      "command": "npx",
      "args": [
        "datadog-mcp-server",
        "--apiKey",
        "<YOUR_API_KEY>",
        "--appKey",
        "<YOUR_APP_KEY>",
        "--site",
        "<YOUR_DD_SITE>",
        "--logsSite",
        "<YOUR_LOGS_SITE>",
        "--metricsSite",
        "<YOUR_METRICS_SITE>"
      ]
    }
  }
}

Locations for the Claude Desktop config file:

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

Usage with MCP Inspector

To use with the MCP Inspector tool:

npx @modelcontextprotocol/inspector datadog-mcp-server --apiKey=your_api_key --appKey=your_app_key

Available Tools

The server provides these MCP tools:

get-monitors: Fetch monitors with optional filtering
get-monitor: Get details of a specific monitor by ID
get-dashboards: List all dashboards
get-dashboard: Get a specific dashboard by ID
get-metrics: List available metrics
get-metric-metadata: Get metadata for a specific metric
get-events: Fetch events within a time range
get-incidents: List incidents with optional filtering
search-logs: Search logs with advanced query filtering
aggregate-logs: Perform analytics and aggregations on log data

Examples

Example: Get Monitors

{
  "method": "tools/call",
  "params": {
    "name": "get-monitors",
    "arguments": {
      "groupStates": ["alert", "warn"],
      "limit": 5
    }
  }
}

Example: Get a Dashboard

{
  "method": "tools/call",
  "params": {
    "name": "get-dashboard",
    "arguments": {
      "dashboardId": "abc-def-123"
    }
  }
}

Example: Search Logs

{
  "method": "tools/call",
  "params": {
    "name": "search-logs",
    "arguments": {
      "filter": {
        "query": "service:web-app status:error",
        "from": "now-15m",
        "to": "now"
      },
      "sort": "-timestamp",
      "limit": 20
    }
  }
}

Example: Aggregate Logs

{
  "method": "tools/call",
  "params": {
    "name": "aggregate-logs",
    "arguments": {
      "filter": {
        "query": "service:web-app",
        "from": "now-1h",
        "to": "now"
      },
      "compute": [
        {
          "aggregation": "count"
        }
      ],
      "groupBy": [
        {
          "facet": "status",
          "limit": 10,
          "sort": {
            "aggregation": "count",
            "order": "desc"
          }
        }
      ]
    }
  }
}

Example: Get Incidents

{
  "method": "tools/call",
  "params": {
    "name": "get-incidents",
    "arguments": {
      "includeArchived": false,
      "query": "state:active",
      "pageSize": 10
    }
  }
}

Troubleshooting

If you encounter a 403 Forbidden error, verify that:

Your API key and Application key are correct
The keys have the necessary permissions to access the requested resources
Your account has access to the requested data
You're using the correct endpoint for your region (e.g., datadoghq.eu for EU customers)

Debugging

If you encounter issues, check Claude Desktop's MCP logs:

# On macOS
tail -n 20 -f ~/Library/Logs/Claude/mcp*.log

# On Windows
Get-Content -Path "$env:APPDATA\Claude\Logs\mcp*.log" -Tail 20 -Wait

Common issues:

403 Forbidden: Authentication issue with Datadog API keys
API key or App key format invalid: Ensure you're using the full key strings
Site configuration errors: Make sure you're using the correct Datadog domain
Endpoint mismatches: Verify that service-specific endpoints are correctly set if you're using separate domains for logs and metrics

License

MIT

Datadog