A Model Context Protocol (MCP) server that enables AI assistants (GitHub Copilot, Claude, etc.) to interact with Datadog's observability platform. Query metrics, search logs, trace APM data, manage monitors/alerts, analyze dashboards, and investigate incidents — all through natural language.

Monorepo Structure

packages/
  datadog-server/          # MCP server (16 tools)
  datadog-vscode-extension/ # VS Code extension

Quick Start

As a standalone MCP server

# Install
npm install -g @fazorboy/datadog-mcp-server

# Set environment variables
export datadog_DD_API_KEY="your-api-key"
export datadog_DD_APP_KEY="your-app-key"

# Run
datadog-mcp-server

As a VS Code extension

Install the VSIX from releases, set your API keys in Settings > Datadog MCP, and the server appears in your MCP Servers panel automatically.

Documentation

Server Documentation — All 16 tools, configuration, API reference
VS Code Extension — Installation, configuration, usage

Usage Examples

Once the MCP server is running (standalone or via the VS Code extension), ask your AI assistant:

Prompt	Tool Used
"Show me CPU metrics for `host:web-prod-01` over the last 4 hours"	`datadog_query_metrics`
"Search logs for errors containing 'timeout' in service `payment-api` from the past hour"	`datadog_search_logs`
"List all monitors that are currently in alert state"	`datadog_list_monitors`
"Create a monitor that alerts when `avg:system.cpu.user{env:prod}` exceeds 90% for 5 minutes"	`datadog_create_monitor`
"Show me the APM traces for slow requests (>2s) in `checkout-service`"	`datadog_search_traces`
"What services depend on `postgres-primary`? Show me the service map"	`datadog_service_map`
"List all dashboards matching 'infrastructure'"	`datadog_list_dashboards`
"Get the widgets and layout for dashboard `abc-def-123`"	`datadog_get_dashboard`
"Show me all hosts with the tag `env:staging` and their load averages"	`datadog_list_hosts`
"Mute the 'High Error Rate' monitor for 2 hours for a deploy"	`datadog_mute_monitor`

License

MIT

Datadog MCP Server