Datadog MCP Server (Official Remote)
OfficialDatadog's managed remote server: query logs, metrics, traces, monitors and incidents.
Add to your client
Copy the config for your MCP client and paste it into its config file.
claude mcp add --transport http datadog https://mcp.datadoghq.com/api/unstable/mcp-server/mcpPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"datadog-mcp-server-official-remote": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.datadoghq.com/api/unstable/mcp-server/mcp"
]
}
}
}Claude Desktop connects to remote servers through the `mcp-remote` proxy (installed on first run via npx). Restart Claude Desktop after saving.
Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- A Datadog account on a supported site (e.g. US1 or EU), with access to the MCP server Preview enabled for your org
- An MCP client that supports remote HTTP servers with OAuth (e.g. Claude Code, Cursor, OpenAI Codex CLI)
- Authorization to your Datadog org — the remote server uses an OAuth sign-in flow; Datadog API and application keys are used for the alternative/self-hosted auth path
- Appropriate Datadog roles/permissions for the data you want to query (logs, APM, incidents, etc.)
About Datadog MCP Server (Official Remote)
This is Datadog's official, Datadog-hosted remote MCP server (a.k.a. the Bits AI MCP server). It connects AI agents and coding assistants directly to your Datadog organization so they can query observability data — logs, metrics, traces and spans, monitors, incidents, dashboards, and host/infrastructure context — in natural language.
Because it is a managed remote server, there is no local process to run: you point an MCP-capable client at the Datadog endpoint for your site and authorize it. Datadog built it in collaboration with OpenAI (it ships in the OpenAI Codex CLI) and it also works with Claude Code, Cursor, and other standards-compliant MCP clients.
The server handles Datadog API communication and returns structured results with clear errors, letting an agent investigate an incident, pull the timeseries behind an alerting monitor, or correlate logs and traces during debugging. At the time of writing it is offered in Preview and you request access through Datadog's documentation.
Tools & capabilities (11)
search_datadog_logsSearch and retrieve logs by query, facets, tags, and time range
get_logs / query log analyticsAggregate and analyze logs (grouping, stats) over a time window
query_timeseries_dataQuery metric timeseries for a metric query over a start/end time range
list_metricsDiscover available metrics and their metadata/context
search_datadog_spans / list_spansSearch APM distributed spans for tracing investigations
get_traceFetch a complete distributed trace by ID
list_monitorsList monitors with their configs, thresholds, and alert states (filter by name/tag/state)
get_monitorGet details for a specific monitor
list_incidents / get_incidentList active incidents and retrieve a specific incident's details
list_dashboardsDiscover available dashboards in your org with context
list_hostsList detailed host / infrastructure information
When to use it
- Use it when debugging an incident and you want the agent to pull related logs, traces, and the firing monitor together
- Use it when you want to ask 'what's the p99 latency of service X over the last hour' in plain English and get the timeseries
- Use it when triaging alerts and need a monitor's configuration, thresholds, and current state without leaving your editor
- Use it when correlating a deploy with a spike by searching spans and logs around a time window
- Use it when you want incident context (active incidents, details) surfaced inside Claude Code, Cursor, or Codex
Quick setup
- 1Request access to the Datadog MCP server Preview for your org in Datadog's documentation
- 2Add a remote MCP server entry using the endpoint for your site (US1: https://mcp.datadoghq.com/api/unstable/mcp-server/mcp, EU: https://mcp.datadoghq.eu/api/unstable/mcp-server/mcp)
- 3Authorize the connection via the OAuth sign-in flow when your client prompts (or configure API + app keys for the alternative auth path)
- 4Restart your MCP client and confirm the Datadog tools are listed
- 5Run a query (e.g. list recent monitors or search logs) to verify access and permissions
Security notes
This is an unstable/preview API endpoint that authenticates against your Datadog org via browser flow, so it can read sensitive logs and traces. Restrict which users can authorize it and prefer read-only scopes.
Datadog MCP Server (Official Remote) FAQ
Is this the official Datadog MCP server?
Yes — it is built and hosted by Datadog (the Bits AI MCP server), developed in collaboration with OpenAI, and is distinct from community Datadog MCP servers.
How do I authenticate?
The managed remote server uses an OAuth browser sign-in to your Datadog org; an alternative path uses Datadog API and application keys. You must have permissions for the data you query.
Which endpoint do I use?
Use the endpoint for your Datadog site — US1 is https://mcp.datadoghq.com/api/unstable/mcp-server/mcp and EU is https://mcp.datadoghq.eu/api/unstable/mcp-server/mcp, with regional variants for other sites.
Which clients are supported?
Any standards-compliant MCP client that supports remote HTTP servers, including Claude Code, Cursor, and the OpenAI Codex CLI.
Is it generally available?
It is offered in Preview at the time of writing; you request access through Datadog's documentation, and the API path is marked 'unstable' in the endpoint.
Alternatives to Datadog MCP Server (Official Remote)
Compare all alternatives →
Official Elastic server: list indices, read mappings, and search with Query DSL.
Official PostHog server: product analytics, feature flags, experiments, error tracking and SQL.
Run PromQL queries and analyze Prometheus metrics from any MCP client.
Compare Datadog MCP Server (Official Remote) with: