A Datadog MCP server lets an AI query your observability data — metrics, logs, monitors, dashboards, APM traces and security signals — directly from a chat. Datadog maintains an official remote server, so setup is an OAuth login rather than a config file. Here's how to connect it.
The pick: Datadog's official remote server
Datadog's MCP server is hosted and remote, with OAuth auth and in-product visualisations (MCP Apps). It exposes most Datadog products read-side, which is exactly what you want for incident triage and root-cause questions.
Option A — official connector (recommended)
In Claude, click the + under the prompt → Add Connector, find Datadog in the directory, enable it, and complete the OAuth login for your Datadog site.
Option B — Claude Code
For first-time setup in Claude Code, run:
/ddsetup
…or just enter any Datadog-related prompt. Select your Datadog site/region and complete the OAuth login when prompted.
If the directory connector isn't available to you, add Datadog as a custom connector using the remote MCP URL for your site (the host differs by region — check Datadog's docs for us1, eu, etc.).
Verify
Ask: "Show error-rate metrics for the checkout service over the last hour and any triggered monitors." Live data back confirms it works.
Common problems
- Wrong region — Datadog's MCP URL is site-specific; an
euorg won't authorise against aus1endpoint. - Empty results — your Datadog role can't see that service/dashboard; access mirrors your own permissions.
- Connector missing — use the custom remote-URL route instead of the directory entry.
Security
The server reads sensitive operational data, so connect with a least-privilege Datadog role and treat OAuth tokens like credentials. See MCP security best practices and real-time agent monitoring.
Datadog vs Sentry
For error tracking specifically, compare Sentry MCP vs Datadog MCP. Datadog is the broader observability platform; Sentry is sharper on exceptions.
Going further
See the Datadog agent profile, Grafana MCP setup, the DevOps category, and the DevOps/SRE loadout.