Dynatrace MCP Server

<h4 align="center">
<a href="https://github.com/dynatrace-oss/dynatrace-mcp/releases">

</a>
<a href="https://github.com/dynatrace-oss/dynatrace-mcp/blob/main/LICENSE">

</a>
<a href="https://vscode.dev/redirect/mcp/install?name=dynatrace-mcp-server&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22%40dynatrace-oss%2Fdynatrace-mcp-server%22%5D%2C%22env%22%3A%7B%7D%7D">

</a>
<a href="https://vscode.dev/redirect/mcp/install?name=dynatrace-mcp-server&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22%40dynatrace-oss%2Fdynatrace-mcp-server%22%5D%2C%22env%22%3A%7B%7D%7D">

</a>
<a href="https://www.npmjs.com/package/@dynatrace-oss/dynatrace-mcp-server">

</a>
<a href="https://github.com/dynatrace-oss/dynatrace-mcp">

</a>
<a href="https://github.com/dynatrace-oss/dynatrace-mcp">

</a>
</h4>

🛠️ MAINTENANCE MODE

Please note, that this repository is currently in _Maintenance Mode_. We will continue fixing/patching issues when they come up.

Recommended Actions

> 🚀 Try Dynatrace-for-Ai and dtctl (recommended for local development use cases like VS Code, IntelliJ, Claude Code, Cursor and alike)

> Try out our new Remote Dynatrace MCP Server! Now available - no local setup required, connect instantly to your Dynatrace environment from any MCP-compatible client. See our migration guide for a comparison and step-by-step instructions. This setup is recommended when connecting another Agent (e.g., Atlassian Rovo, GitHub Coding Agent and alike) to Dynatrace.

The local _Dynatrace MCP server_ allows AI Assistants to interact with the Dynatrace observability platform,
bringing real-time observability data directly into your development workflow.

> Note: This product is not officially supported by Dynatrace.

If you need help, please contact us via GitHub Issues if you have feature requests, questions, or need help.

https://github.com/user-attachments/assets/25c05db1-8e09-4a7f-add2-ed486ffd4b5a

Prerequisites

- Node.js v24 or newer is required to run the Dynatrace MCP Server.

Quickstart

You can add this MCP server to your MCP Client like VSCode, Claude, Cursor, Amazon Q, Windsurf, ChatGPT, or Github Copilot via the command is npx -y @dynatrace-oss/dynatrace-mcp-server (type: stdio). For more details, please refer to the configuration section below.

Furthermore, you need to configure the URL to a Dynatrace environment:

- DT_ENVIRONMENT (string, e.g., https://abc12345.apps.dynatrace.com) - URL to your Dynatrace Platform (do not use Dynatrace classic URLs like abc12345.live.dynatrace.com)

Authentication will be handled via Authorization Code Flow in your browser, you don't need to define a Platform Token nor an OAuth Client to get started. After the initial authentication, your token is securely stored in the OS keychain (macOS Keychain, Windows Credential Manager, or Linux Secret Service) and reused on subsequent server starts, so the browser only opens once per token lifetime.

If you are running in a headless/container environment where the OS keychain is unavailable, set DT_MCP_TOKEN_STORAGE=file to persist tokens in ~/.config/dynatrace-mcp/ instead.

Once you are done, we recommend looking into example prompts, like Get all details of the entity 'my-service' or Show me error logs. Please mind that these prompts lead to executing DQL statements which may incur costs in accordance to your licence.

VSCode

{
  "servers": {
    "npx-dynatrace-mcp-server": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT": "https://abc12345.apps.dynatrace.com"
      }
    }
  }
}

Architecture

Use cases

- Real-time observability - Fetch production-level data for early detection and proactive monitoring
- Contextual debugging - Fix issues with full context from monitored exceptions, logs, and anomalies
- Security insights - Get detailed vulnerability analysis and security problem tracking
- Natural language queries - Use AI-powered DQL generation and explanation
- Multi-phase incident investigation - Systematic 4-phase approach with automated impact assessment
- Advanced transaction analysis - Precise root cause identification with file/line-level accuracy
- Cross-data source correlation - Connect problems → spans → logs with trace ID correlation
- DevOps automation - Deployment health gates with automated promotion/rollback logic
- Security compliance monitoring - Multi-cloud compliance assessment with evidence-based investigation

Capabilities

- Observability & Problem Management (e.g., list_problems, list_vulnerabilities, list_exceptions, get_kubernetes_events)
- Querying Data from Grail (e.g., execute_dql, verify_dql, generate_dql_from_natural_language, explain_dql_in_natural_language)
- Entity Discovery (e.g., find_entity_by_name)
- Dynatrace Intelligence (e.g., chat_with_davis_copilot, list_davis_analyzers, execute_davis_analyzer)
- Automation & Notifications (e.g., send_slack_message, send_email, send_event)
- Share results (e.g., create_dynatrace_notebook)

Costs

Important: While this local MCP server is provided for free, using certain capabilities to access data in Dynatrace Grail may incur additional costs based
on your Dynatrace consumption model. This affects execute_dql tool and other capabilities that query Dynatrace Grail storage, and costs
depend on the volume (GB scanned).

Before using this MCP server extensively, please:

1. Review your current Dynatrace consumption model and pricing
2. Understand the cost implications of the specific data you plan to query (logs, events, metrics) - see Dynatrace Pricing and Rate Card
3. Start with smaller timeframes (e.g., 12h-24h) and make use of buckets to reduce the cost impact
4. Set an appropriate DT_GRAIL_QUERY_BUDGET_GB environment variable (default: 1000 GB) to control and monitor your Grail query consumption

Grail Budget Tracking:

The MCP server includes built-in budget tracking for Grail queries to help you monitor and control costs:

- Set DT_GRAIL_QUERY_BUDGET_GB (default: 1000 GB) to define your session budget limit
- The server tracks bytes scanned across all Grail queries in the current session
- You'll receive warnings when approaching 80% of your budget
- Budget exceeded alerts help prevent unexpected high consumption
- Budget resets when you restart the MCP server session

To understand costs that occured:

Execute the following DQL statement in a notebook to see how much bytes have been queried from Grail (Logs, Events, etc...):

fetch dt.system.events
| filter event.kind == "QUERY_EXECUTION_EVENT" and contains(client.client_context, "dynatrace-mcp")
| sort timestamp desc
| fields timestamp, query_id, query_string, scanned_bytes, table, bucket, user.id, user.email, client.client_context
| maketimeSeries sum(scanned_bytes), by: { user.email, user.id, table }

AI-Powered Assistance

- Natural Language to DQL - Convert plain English queries to Dynatrace Query Language
- DQL Explanation - Get plain English explanations of complex DQL queries
- AI Chat Assistant - Get contextual help and guidance for Dynatrace questions

Configuration

You can add this MCP server (using STDIO) to your MCP Client like VS Code, Claude, Cursor, Amazon Q Developer CLI, Windsurf Github Copilot via the package @dynatrace-oss/dynatrace-mcp-server.

We recommend to always set it up for your current workspace instead of using it globally.

VS Code

{
  "servers": {
    "npx-dynatrace-mcp-server": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT": "https://abc12345.apps.dynatrace.com"
      }
    }
  }
}

Claude Desktop

{
  "mcpServers": {
    "dynatrace-mcp-server": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT": "https://abc12345.apps.dynatrace.com"
      }
    }
  }
}

Amazon Q Developer CLI

The Amazon Q Developer CLI provides an interactive chat experience directly in your terminal. You can ask questions, get help with AWS services, troubleshoot issues, and generate code snippets without leaving your command line environment.

{
  "mcpServers": {
    "dynatrace-mcp-server": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT": "https://abc12345.apps.dynatrace.com"
      }
    }
  }
}

This configuration should be stored in <your-repo>/.amazonq/mcp.json.

Amazon Kiro

The Amazon Kiro is an agentic IDE that helps you do your best work with features such as specs, steering, and hooks.

{
  "mcpServers": {
    "dynatrace-mcp-server": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT": "https://abc12345.apps.dynatrace.com"
      }
    }
  }
}

This configuration should be stored in <your-repo>/.kiro/settings/mcp.json.

Google Gemini CLI

The Google Gemini CLI is Google's official command-line AI assistant that supports MCP server integration. You can add the Dynatrace MCP server using either the built-in management commands or manual configuration.

Using gemini CLI directly (recommended):

gemini extensions install https://github.com/dynatrace-oss/dynatrace-mcp
export DT_PLATFORM_TOKEN=... # optional

The command will ask for the value Dynatrace Environment.

Verify that the server is running via

gemini mcp list

Or manually in your ~/.gemini/settings.json or .gemini/settings.json:

{
  "mcpServers": {
    "dynatrace": {
      "command": "npx",
      "args": ["@dynatrace-oss/dynatrace-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT": "https://abc12345.apps.dynatrace.com"
      },
      "timeout": 30000,
      "trust": false
    }
  }
}

HTTP Server Mode (Alternative)

> 💡 Recommendation: Instead of self-hosting with --http / --server, consider using the official Dynatrace Remote MCP Server — it is more secure, always up to date, and requires no infrastructure management.

For scenarios where you need to run the MCP server as an HTTP service instead of using stdio (e.g., for stateful sessions, load balancing, or integration with web clients), you can use the HTTP server mode:

Running as HTTP server:

Please secure your server with a bearer token:

export MCP_BEARER_TOKEN=<bearer-token>

```bash