Datadog MCP Server – Observability and Incident Insights
Datadog's official managed MCP server gives AI agents structured access to observability and operational data across supported Datadog products. Use it for production investigation, monitoring analysis, incident response, performance troubleshooting, and related engineering workflows.
Overview
Datadog's official managed MCP server connects MCP-compatible AI agents to live
Datadog telemetry and platform context. It lets an agent retrieve structured
information from supported Datadog products instead of relying only on copied
dashboards, screenshots, stack traces, or static model knowledge.
What the MCP server enables
Datadog organizes tools into product-specific toolsets. Depending on the
authenticated user's permissions and the toolsets enabled in the connection,
an AI agent can work with areas such as:
- Logs, metrics, traces, services, and APM data.
- Monitors, dashboards, notebooks, incidents, and cases.
- Security signals and supported cloud-security context.
- Software Delivery, CI Visibility, and test-optimization data.
- LLM Observability traces, spans, and experiment results.
- Feature Flags and other supported Datadog platform capabilities.
- Datadog documentation and operational guidance.
Tool responses are designed for context efficiency. Many tools support a
max_tokens parameter, and Datadog may truncate large responses with guidance
on how the agent can request additional details.
When to use it
Use Datadog MCP when an AI agent needs current production context to investigate
an issue or support an operational workflow. Typical examples include finding
errors associated with a deployment, analyzing a latency regression, reviewing
logs and traces during an incident, summarizing monitor activity, correlating
security signals, inspecting CI failures, and preparing a troubleshooting
summary for human review.
Connection and authentication
Datadog provides a managed Streamable HTTP endpoint. For the US1 site, the
documented endpoint is
https://mcp.datadoghq.com/api/unstable/mcp-server/mcp. Other Datadog sites use
regional endpoints shown in the Datadog site selector.
Authentication uses an interactive OAuth flow. Compatible clients open a
browser so the user can sign in and authorize access. Product-specific tools
can be enabled with the toolsets query parameter, while individual tools can
be excluded with omit_tools.
Key considerations
Access follows the signed-in user's Datadog roles and permissions. Enable only
the toolsets required for the workflow and review write or high-impact actions
before execution. Datadog records MCP tool calls in Audit Trail, including tool
names, arguments, user identity, and client metadata, and emits usage metrics
for monitoring the integration. The managed server is HIPAA-eligible, but the
connected AI client must independently satisfy applicable compliance needs.
The service is not compatible with Datadog GovCloud, and its tools remain under
active development and may change.
Supported Transports
streamable_http
URL: https://mcp.datadoghq.com/api/unstable/mcp-server/mcp
Frequently Asked Questions
- When should an AI agent use the Datadog MCP server?
- Use it when a workflow needs current Datadog observability or operational context, such as investigating production errors, analyzing latency, reviewing logs and traces, summarizing monitor activity, supporting incident response, inspecting CI failures, or correlating security signals.
- What does the Datadog MCP server add to an AI agent's capabilities?
- It gives the agent structured access to live Datadog telemetry, platform resources, and supported operational tools, allowing it to investigate current systems instead of relying only on static model knowledge or manually pasted dashboards and error data.
- What can an AI agent access or manage through Datadog MCP?
- Depending on enabled toolsets and user permissions, the agent can work with logs, metrics, traces, APM, monitors, dashboards, incidents, cases, security signals, software-delivery data, LLM Observability, feature flags, and related Datadog documentation and platform context.
- How is authentication configured for the Datadog MCP server?
- The managed server uses an interactive OAuth flow. The user signs in through a browser and grants access, after which Datadog applies the user's existing roles and permissions. No static API key or bearer-token header is included in the documented standard configuration.
- Which transport should be used for the Datadog MCP server?
- Use Streamable HTTP with the managed endpoint for the user's Datadog site. The YAML includes the documented US1 endpoint; organizations on another Datadog site should replace it with the regional endpoint shown in Datadog's official setup documentation.