Datadog logo
Developer ToolsAuth: Bearer tokenHosting: Hosted or self-hosted

Datadog MCP Server

Let AI agents query your observability data — metrics, monitors, logs, and incidents.

Updated

What is Datadog MCP?

The Datadog MCP server is a Model Context Protocol server that gives AI agents read access to your Datadog observability data. Connected to it, an agent can search metrics, look up monitors and their status, query logs, pull APM traces, and read incidents — so you can ask questions like "why did latency spike at 2pm?" and have the agent gather the evidence from Datadog itself.

It authenticates with a Datadog API key plus application key and is scoped to your Datadog org and site (US, EU, etc.).

Tools the Datadog MCP server exposes

ToolWhat it does
search_metricsFind metrics by name and query time series
get_monitorsList monitors and their current alert status
search_logsQuery logs with Datadog's log search syntax
get_tracesRetrieve APM traces for a service
list_incidentsList and read incidents and their timelines
get_dashboardsLook up dashboards and their widgets

Connect the Datadog MCP server

Connect Datadog MCP to Claude Code

  1. 1

    Create keys

    In Datadog, create an API key and an application key under Organization Settings.

  2. 2

    Add the server

    Configure the Datadog MCP server with `DD_API_KEY`, `DD_APP_KEY`, and your `DD_SITE` (e.g. `datadoghq.com`).

  3. 3

    Verify

    Ask Claude to "list my Datadog monitors that are alerting" to confirm access.

Scope the application key to read-only and the minimum required Datadog permissions.

Connect Datadog MCP to Cursor

  1. 1

    Open MCP settings

    Add a new MCP server in Cursor's settings.

  2. 2

    Provide keys

    Set the Datadog API key, app key, and site as environment variables for the server.

  3. 3

    Test

    Search a metric from chat to confirm the tools respond.

Sample use cases

An alert fires and you want context fast

The agent reads the monitor, pulls the related metric and logs, and summarizes the likely cause.

Post-incident write-up

The agent reads the incident timeline and drafts a retrospective from the Datadog record.

A daily reliability digest

The agent checks monitor status across services and posts a health summary every morning.

Correlate a deploy with metric regressions

The agent queries time-series data before and after a deploy timestamp and flags any p99 latency or error-rate spikes.

Debug a slow API endpoint using traces

The agent retrieves APM traces for the affected service, identifies the slowest spans, and points to the bottleneck.

Audit monitor coverage across services

The agent lists all monitors, groups them by service tag, and highlights services with no alerting configured.

Generate a weekly SLO/SLI report

The agent pulls key metrics, calculates uptime and error budgets, and formats a stakeholder-ready summary.

Search logs for error patterns after a customer complaint

The agent runs a log query filtered by customer ID and time range, then surfaces the relevant error entries.

Security & permissions

The Datadog MCP server uses an API key and application key; the application key inherits the permissions of the user or service account that created it, so create it under a read-only scope for agent use. Pin the correct `DD_SITE` so requests go to your region. Keys should be stored as secrets, never in a committed config, and rotated from Datadog's key management if exposed.

Using Datadog MCP with Gamut

On Gamut, the Datadog MCP server turns observability into an input for automated agents. Trigger an agent on a PagerDuty alert or a schedule, have it gather the relevant Datadog metrics, monitors, and logs, correlate with other connected systems, and post a written summary to Slack or an incident doc — credentials and scheduling handled by Gamut so the investigation starts before a human even opens the dashboard.

Frequently asked questions

What is the Datadog MCP server?

A Model Context Protocol server from Datadog that lets AI agents query your metrics, monitors, logs, traces, and incidents as tools.

How do I connect the Datadog MCP server?

Create a Datadog API key and application key, set them plus your Datadog site as environment variables for the MCP server, and add it to Claude, Cursor, or Gamut.

Is the Datadog MCP server read-only?

By default, yes — create the application key under a read-only scope for the recommended autonomous-agent setup. Write scopes can be added if you want agents to mute monitors or create dashboards.

Which Datadog site does the MCP server use?

Set DD_SITE to your org's region (for example datadoghq.com for US1 or datadoghq.eu for EU) so requests reach the right endpoint.

Can the agent create or modify monitors?

The default read-only scope cannot. If you grant the application key a write scope, the agent could create or update monitors — but the recommended setup for autonomous use keeps it read-only for safety.

Does the agent have access to all my Datadog data?

Access is scoped by the application key's permissions. You can restrict it to specific services, tags, or data types using Datadog's RBAC roles, so the agent only sees what you allow.

Can I use it with multiple Datadog organizations?

Each MCP server instance connects to one Datadog organization (determined by the API key). To work with multiple orgs, configure separate server instances with different API keys.