Dynatrace Managed MCP Server

Dynatrace Managed is the self-hosted edition of Dynatrace, and it is deliberately not covered by the SaaS MCP server. This one is: it connects an AI assistant to one or more on-premises Dynatrace Managed clusters and brings observability data into the debugging session instead of leaving it in another browser tab.

What it can reach. Problems raised against your services, including Kubernetes workloads. Security problems and vulnerability details, with enough structure for evidence-based investigation. Monitored entities and their relationship mappings. Service Level Objectives with evaluation state and error budgets. System events. Log search with content and time-based filtering. And metric queries through the V2 Metrics API.

Two situations it is built for. Either Dynatrace Managed is still your primary observability system and holds the live data — or you have migrated to Dynatrace SaaS but left historical data behind, in which case this server reads the history while a separate Dynatrace SaaS MCP handles anything recent. Multi-environment support is native: several clusters are declared in one config file and addressed by alias, so an agent can compare production against staging without reconnecting.

Configuration. A dt-config.yaml or dt-config.json listing each environment's apiEndpointUrl, environmentId, alias and API token (token references such as ${DT_PROD_TOKEN} are resolved from the environment), pointed at by DT_CONFIG_FILE. Container deployments can pass the whole list as JSON in DT_ENVIRONMENT_CONFIGS instead. The API token needs read scopes for entities, events, logs, metrics, problems, security problems and SLOs, plus DataExport.

Runs in local stdio mode for a workstation or remote mode over HTTP/SSE for shared and production-like setups. Published under the dynatrace-oss organisation as a community-supported open source project — feature requests and questions go through GitHub Issues, not Dynatrace support. Apache-2.0 licensed.