Dynatrace MCP

The Dynatrace MCP server connects an AI client to a Dynatrace environment so an agent can read observability data directly instead of working from guesses. It exposes tools that generate and explain Dynatrace Query Language, then execute those queries against the environment to return logs, metrics, spans, events, entities and Smartscape topology. Other tools pull open problems and security vulnerabilities, analyze Kubernetes events, forecast time series, resolve entity names to internal IDs, and search Dynatrace documentation and troubleshooting guides. Access runs through the MCP gateway endpoint hosted on the Dynatrace environment itself, so there is no local install to maintain. Authentication uses either a platform token or an OAuth client with the authorization code grant, which supports automatic token refresh. The token carries the scopes that decide what an agent may touch: permission to read and invoke gateway servers, permission to run natural language to DQL translation, storage read permissions, and any further scope a workflow needs. Because the bearer token is bound to a user, the agent sees exactly the data that person can already see and nothing beyond it. It suits SREs, platform engineers and application developers who want to triage an incident from an IDE or chat client: ask why a service degraded, have the agent write and run the DQL, then follow the entity graph into the failing component without opening the console.