Datadog MCP. Query metrics, search logs, and check alerts from chat.

You're connecting your AI client straight into your infrastructure monitoring stack. This isn't some dashboard you gotta open up and click through; it's direct, conversational access to real-time application health data. You can query performance metrics, search logs for errors, and check system monitor status without the usual headache of building complex queries.

Checking System Health Status

The list_monitors tool lets you quickly see what's running rough right now. It filters all your configured monitors by their operational state—you can immediately pull up a list showing only alerts, warnings, or green 'ok' statuses. You don't have to scroll through hundreds of rules just to find the broken ones.

Once you identify an issue using list_monitors, you can drill down with the get_monitor tool. This pulls all the nitty-gritty details for a single monitor ID, including its current notification settings, what its specific threshold values are set to, and a full history of every status change it’s reported over time.

If an alert is annoying you right now but you know it'll clear up in twenty minutes, you can use mute_monitor to temporarily silence the alarm for a fixed amount of time.

Analyzing Performance Trends Over Time

The query_metrics tool lets you pull time-series data—think CPU usage or total request counts—across multiple dimensions. You specify a historical timeframe using defined UNIX timestamps, and it spits out the raw performance data for that window. This helps you track patterns like load spikes or latency creep over days or weeks.

To get a bird's-eye view of your metrics setup, use get_dashboard. It retrieves all widget configurations, template variables, and the whole layout structure for any dashboard visualization, letting your AI client know exactly what data points are available to analyze. Meanwhile, if you just need to see what dashboards exist in the system—maybe you can't find where that key metric is displayed—list_dashboards shows all available boards by title, type (like a timeboard or screenboard), and even gives you the direct access URL for each one.

Finding Specific Errors in Application Logs

The search_logs tool lets your agent search through massive collections of structured application logs using query syntax. You can find entries matching specific error codes, user IDs, or particular service names by specifying the criteria. The results return timestamps and status levels right alongside the log entry.

For a broader view of what's happening system-wide, list_events gathers recent system events—this includes everything from high-priority alerts to successful deployment completions—along with their defined priority level. If you need to know about planned outages so you don't freak out when the system goes down, list_downtimes returns known maintenance periods, including scope tags and the current status of that scheduled work.

Auditing Infrastructure Metadata and Compliance

You can audit every machine connected by using the list_hosts tool. This gives you metadata for every monitored host, showing things like which agent version they're running or what cloud provider they report to—super useful if you gotta check compliance across multiple environments.

To see how well your services are actually performing against internal goals, use list_slos. It retrieves Service Level Objective definitions, letting you check target percentages and the current compliance status for every service you've defined. Finally, you can get a quick rundown of all active monitors using list_monitors to keep track of what needs attention.