Dynatrace for AI
Everything AI agents need to work with Dynatrace, starting with skills.
Skills are portable knowledge packages following the Agent Skills open format. They give AI coding agents the domain-specific context to query, analyze, and interpret Dynatrace data. They work with Claude Code, GitHub Copilot, Cursor, OpenCode, Gemini CLI, and 30+ other compatible tools.
Installation
Skills Package (skills.sh)
npx skills add dynatrace/dynatrace-for-ai
Works with Claude Code, Cursor, Cline, GitHub Copilot, OpenCode, and other compatible agents.
Claude Code Plugin
claude plugin install dynatrace@claude-plugins-official
Prefer browsing interactively? Run /plugin in Claude Code, open the Discover tab, and search "Dynatrace".
The plugin includes skills as well as the Dynatrace MCP server. Set these environment variables before starting Claude Code to connect it to your environment:
export DT_ENVIRONMENT=https://<env>.apps.dynatrace.com # e.g. https://abc12345.apps.dynatrace.com
export DT_PLATFORM_TOKEN=<your-platform-token>
Verify that it works via:
claude mcp login plugin:dynatrace:dynatrace
Please consult the Dynatrace MCP server docs for a full list of scopes required for using the MCP Server with a Platform Token.
To update the plugin, use:
claude plugin marketplace update
claude plugin update dynatrace@claude-plugins-official
Note: The bundled MCP server connects to your Dynatrace environment. See the Dynatrace Privacy Policy for details on data handling.
Cursor Plugin
The Dynatrace plugin is listed on the Cursor Marketplace. Install is GUI-only: open Customize in the Cursor sidebar, find "Dynatrace", select Install, and choose project or user scope.
Set the same DT_ENVIRONMENT and DT_PLATFORM_TOKEN environment variables shown above before starting Cursor.
Manual
Copy skill directories into your agent's skills path (.agents/skills/, .claude/skills/, .cursor/skills/, etc.).
Connecting to Dynatrace
Skills provide knowledge only. To run live queries and manage your environment, pair them with a tool.
Dynatrace CLI (dtctl)
dtctl is a kubectl-style CLI for the Dynatrace platform. It ships with its own Agent Skill that teaches agents how to operate it.
brew install dynatrace-oss/tap/dtctl # Install
dtctl auth login --context my-env \
--environment "https://<env>.apps.dynatrace.com" # Authenticate
npx skills add dynatrace-oss/dtctl # Install the dtctl skill
dtctl doctor # Verify setup
Or install the dtctl skill with dtctl itself: dtctl skills install
Dynatrace MCP Server
The Dynatrace MCP server provides Dynatrace API access via MCP. The Claude Code plugin bundles the MCP server configuration automatically — just set DT_ENVIRONMENT and DT_PLATFORM_TOKEN as shown above. For other agents that support MCP natively, see the MCP server docs.
The MCP server is remote — Dynatrace hosts it, so there is nothing to install. Releases are published to the official MCP registry as io.github.Dynatrace/dynatrace-for-ai, so MCP hosts that read the registry can add it by name.
Skills
DQL & Query Language
Observability
Security
Instrumentation Setup
Platform
Migration
Prompts
Prompts are reusable task templates for common Dynatrace workflows. You can copy them from the /prompts/ directory and paste them directly into any AI chat. For VS Code/GitHub Copilot users, copy prompts into .github/prompts/ to use as slash commands (e.g. /troubleshoot-problem).
Each prompt references the relevant skills above — load those skills first for best results.
How Skills Work
Skills follow the Agent Skills specification and use progressive disclosure:
- Catalog - Agents load only
name+description(~100 tokens per skill) to know what's available. - Instructions - When relevant, the full
SKILL.mdis loaded (<5000 tokens). - Resources - Detailed reference files in
references/are loaded on demand.
Install all skills without penalty. Agents only load what they need.
Contributing
Skill content (skills/) is maintained internally at Dynatrace and published here periodically. PRs that modify skill files will not be accepted — please open an issue instead to suggest changes. See CONTRIBUTING.md for details on what can be contributed directly.
License
Apache-2.0
