Back to MCP Servers

Datadog MCP Server

Datadog's official remote MCP server. Gives AI agents secure, real-time access to logs, metrics, traces, monitors, incidents, dashboards, and more.

Developer Tools by Datadog OAuth2 active
Connect (Remote Server) Docs
Overview

The Datadog MCP Server is an officially hosted remote Model Context Protocol server that bridges AI coding agents and DevOps tools to live Datadog observability data. It exposes a large surface area of Datadog functionality across logs, APM traces, metrics, monitors, incidents, dashboards, security signals, RUM, Database Monitoring, Synthetic tests, CI/CD events, feature flags, Kubernetes resources, and workflows. The server handles authentication, endpoint routing, and tool selection automatically so agents like Claude Code, Cursor, OpenAI Codex, and custom agents can answer natural language questions against production telemetry.

Capabilities are organized into toolsets that can be enabled via a toolsets query parameter on the endpoint URL. Available toolsets include core, alerting, apm, cases, dashboards, dbm, ddsql, error-tracking, feature-flags, kubernetes, llmobs, networks, onboarding, product-analytics, reference-tables, security, software-delivery, synthetics, and workflows. Setting toolsets=all enables every generally available toolset. The default core toolset alone covers searching logs, querying metrics, fetching traces, listing monitors and incidents, and creating notebooks.

Notable points: the server is HIPAA-eligible, respects existing Datadog RBAC (requires mcp_read and/or mcp_write permissions), and supports all major commercial Datadog sites (US1, US3, US5, EU1, AP1, AP2). It is not supported on GovCloud sites. Usage is rate limited to 50 tool calls per 10 second burst, 5,000 calls per day, and 50,000 calls per month.

Tools

Tool Description
search_datadog_logs Search logs with filters (time, query, service, host, storage tier).
analyze_datadog_logs Analyze logs using SQL queries for aggregations and statistical analysis.
get_datadog_metric Query and analyze historical or real-time metric data with custom queries.
search_datadog_metrics List available metrics with filtering and metadata.
get_datadog_trace Fetch a complete APM trace by trace ID.
search_datadog_spans Retrieve APM spans with filters such as service, time, resource.
search_datadog_monitors Retrieve Datadog monitors including statuses and thresholds.
create_datadog_monitor Create a Datadog monitor in draft mode.
validate_datadog_monitor Validate a monitor definition before creating or updating.
search_datadog_incidents Retrieve incidents including state, severity, and metadata.
get_datadog_incident Retrieve detailed information about a specific incident.
search_datadog_dashboards List available dashboards and key details.
get_datadog_dashboard Retrieve a dashboard by ID with title, description, tags, widgets.
upsert_datadog_dashboard Create or update a Datadog dashboard.
search_datadog_services List services in the Software Catalog.
search_datadog_hosts List monitored hosts with filtering and search.
search_datadog_security_signals Search and retrieve security signals from Datadog Security.
analyze_security_findings Analyze security findings using SQL queries.
apm_latency_bottleneck_analysis Analyze latency bottlenecks across traces in an anomaly period.
apm_trace_summary Generate an AI-powered summary of a trace.
ddsql_run_query Run a DDSQL query and return results.
search_datadog_error_tracking_issues Search Error Tracking issues across RUM, Logs, and Traces sources.
list_datadog_feature_flags List feature flags with pagination.
search_datadog_k8s_resources Search Kubernetes resources across clusters.
get_synthetics_tests Search Datadog Synthetic tests.
execute_datadog_workflow Execute a published workflow with an agent trigger.
search_datadog_cases Search Case Management cases by status, priority, and other filters.
search_datadog_ci_pipeline_events Search CI pipeline events with filters.
get_datadog_flaky_tests Search Test Optimization for flaky tests.
find_datadog_database_instances Discover and rank database instances for DBM investigation.
Setup Guide

Prerequisites

  • A Datadog account on a supported site (US1, US3, US5, EU1, AP1, or AP2). GovCloud is not supported.
  • Datadog user permissions: mcp_read for reads and mcp_write for create/update tools. The Standard role includes both. Resource-specific permissions (for example "Monitors Read") still apply.
  • An MCP-compatible client: Claude Desktop, Claude Code, Cursor, VS Code (Datadog extension), OpenAI Codex CLI, or Block's Goose.

Endpoint URLs by site

Use the endpoint that matches your Datadog site:

  • US1: https://mcp.datadoghq.com/api/unstable/mcp-server/mcp
  • US3: https://mcp.us3.datadoghq.com/api/unstable/mcp-server/mcp
  • US5: https://mcp.us5.datadoghq.com/api/unstable/mcp-server/mcp
  • EU1: https://mcp.datadoghq.eu/api/unstable/mcp-server/mcp
  • AP1: https://mcp.ap1.datadoghq.com/api/unstable/mcp-server/mcp
  • AP2: https://mcp.ap2.datadoghq.com/api/unstable/mcp-server/mcp

Append ?toolsets=all to enable every available toolset, or pick specific ones, for example ?toolsets=core,apm,security.

Claude Code configuration

{
  "mcpServers": {
    "datadog": {
      "type": "http",
      "url": "https://mcp.datadoghq.com/api/unstable/mcp-server/mcp?toolsets=all"
    }
  }
}

OAuth login is triggered in a browser the first time the client connects.

API key authentication (alternative)

If you cannot use OAuth, pass Datadog API and Application keys as HTTP headers:

{
  "mcpServers": {
    "datadog": {
      "type": "http",
      "url": "https://mcp.datadoghq.com/api/unstable/mcp-server/mcp?toolsets=all",
      "headers": {
        "DD_API_KEY": "<YOUR_API_KEY>",
        "DD_APPLICATION_KEY": "<YOUR_APPLICATION_KEY>"
      }
    }
  }
}

Other clients

  • Claude Desktop: Add a custom connector pointing at the remote MCP URL.
  • Cursor: Install the Datadog plugin from the Cursor Marketplace, then run /ddsetup in agent chat.
  • VS Code: Install the Datadog extension and restart the IDE.
  • Devin: Settings → MCP Marketplace → Datadog, then /ddsetup.

Testing

npx @modelcontextprotocol/inspector

Connect using Streamable HTTP transport to verify tool availability.

Use Cases
  • Investigate a production incident by correlating error logs, APM traces, and active monitors without leaving the IDE.
  • Generate or update Datadog dashboards and monitors from a natural language description, then validate the spec before saving.
  • Triage security signals and run SQL aggregations over security findings to identify high risk patterns.
  • Diagnose database performance regressions with PostgreSQL explain plans, query samples, and DBM recommendations.
  • Audit CI pipeline reliability and identify flaky tests as part of a code review or release workflow.
Example Prompts
  • "Search Datadog logs for 5xx errors on the checkout service in the last 30 minutes and group them by host."
  • "List all monitors currently alerting in production, then open the related incident and summarize what is failing."
  • "Pull the slowest 10 APM traces for service payments-api over the last hour and run a latency bottleneck analysis."
  • "Find Database Monitoring queries on users-db with p95 above 500ms and fetch the explain plan for the worst one."
  • "Create a draft Datadog monitor that alerts when error rate on auth-service exceeds 2% for 5 minutes, then validate it."
Pros
  • Officially built and hosted by Datadog, no self-hosting needed.
  • Very broad coverage across observability domains: logs, metrics, APM, RUM, security, DBM, CI/CD, Kubernetes, and workflows.
  • OAuth-based auth respects existing Datadog RBAC, and the server is HIPAA-eligible.
  • Toolsets query parameter lets you scope which tools the agent sees, keeping context windows manageable.
Limitations
  • Not supported on GovCloud sites (app.ddog-gov.com, us2.ddog-gov.com).
  • Rate limited to 50 calls per 10s, 5,000 per day, and 50,000 per month, which can constrain heavy agent workloads.
  • Still under active development on the api/unstable path, so tool surfaces and behavior may change.
Alternatives
More Developer Tools MCP Servers
AWS API
Bridge that lets AI assistants manage AWS infrastructure by executing validated AWS CLI commands across every AWS service.
AWS CDK
Official AWS Labs MCP server that guides agents through AWS CDK development with prescriptive patterns, CDK Nag security validation, and Solutions Constructs discovery.
AWS CloudWatch
Official AWS Labs MCP server for CloudWatch. Query metrics, analyze logs, investigate alarms, and run PromQL queries for AI-powered root cause analysis.
Aiven
Official MCP server for the Aiven cloud data platform. Manage PostgreSQL, Apache Kafka, and other Aiven services from AI assistants.