Datadog

The Datadog connector lets AI agents query metrics, inspect monitors, browse dashboards, and manage incidents — all via the Datadog REST API. Datadog uses dual API key authentication: an API Key and an Application Key are both required and stored encrypted in MCPGate's vault.

Authentication#

Datadog uses dual API key authentication. To connect:

Sign in to Datadog and navigate to Organization Settings → API Keys
Create or copy an existing API Key and paste it into the first field
Navigate to Organization Settings → Application Keys, create a key, and paste it into the second field
Ensure the user who owns the Application Key has the permissions listed below

Note

Both keys are stored encrypted in MCPGate's vault and are never forwarded to the AI client. The Application Key inherits the permissions of the Datadog user it was created under.

Ensure the user who owns the Application Key has the following permissions.

Permission	Purpose
monitors_read	List and read monitor details
monitors_write	Create, update, and mute monitors
metrics_read	Query metric time series data
dashboards_read	List dashboards
incident_read	List Datadog incidents
incident_write	Create incidents

Tool reference#

The Datadog connector exposes 10 tools. Use the MCP Apps page to enable or disable individual tools per app, and the Guardrails page to add fine-grained rules on top.

Tool	Category	Description
datadog_list_monitors	read	List all Datadog monitors
datadog_get_monitor	read	Get a specific Datadog monitor by ID
datadog_query_metrics	read	Query Datadog metrics for a given time range and query string
datadog_list_dashboards	read	List all Datadog dashboards
datadog_list_incidents	read	List Datadog incidents
datadog_create_monitor	write	Create a new Datadog monitor
datadog_update_monitor	write	Update an existing Datadog monitor
datadog_mute_monitor	write	Mute a Datadog monitor
datadog_create_incident	write	Create a new Datadog incident
datadog_delete_monitor	delete	Delete a Datadog monitor by ID

Common guardrail recipes#

Read-only access#

Enable only the five read tools. Deny all write and delete tools. Useful for metric querying and observability agents that surface dashboard data without modifying monitors.

Operate preset#

Enable read tools plus datadog_mute_monitor and datadog_update_monitor. Deny datadog_create_monitor, datadog_create_incident, and datadog_delete_monitor. This lets the AI silence noisy monitors during maintenance without the ability to add or remove them.

Protect production monitors#

Apply the allow_values template to datadog_delete_monitor and target the monitor_id field. Set an empty allow-list to effectively deny all deletions while keeping other write operations available.

Troubleshooting#

Metrics query syntax — datadog_query_metricsuses Datadog's standard metrics query syntax (e.g. avg:system.cpu.user{*}.rollup(avg, 60)). Invalid syntax returns a 400 error.
Application Key vs. API Key — Some Datadog operations require only the API Key, others require both. If a tool returns 403, verify the Application Key was entered and that the owning user has the required permissions.
Site region — Datadog has regional endpoints (US1, EU1, US3, etc.). MCPGate defaults to the US1 endpoint (api.datadoghq.com). If your organisation uses a different region, contact support.