Dynatrace Managed MCP Server

Latest Dynatrace Managed MCP Server releases

The local Dynatrace Managed MCP server allows AI Assistants to interact with one or more self-hosted Dynatrace Managed deployments, bringing observability data directly into your AI-assisted workflow.

This MCP server supports two modes:

• Local mode: Runs on your machine for development and testing.
• Remote mode: Connects over HTTP/SSE for distributed or production-like setups.

Quickstart in stdio (local) mode

You can add this MCP server to your AI Assistant, such as VSCode, Claude, Cursor, Kiro, Windsurf, ChatGPT, or GitHub Copilot.

To run this MCP server, you to have to configure four things:

• Dynatrace Managed API token
• Configuration file: dt-config.yaml or dt-config.json file which is responsible for defining list of environments you intend to use
• MCP Server connection configuration file: local mcp configuration, which is dependent on tools you are using
• Set DT_CONFIG_FILE to the path of your dt-config.yaml or dt-config.json file in the MCP server environment.

Dynatrace Managed API token

For information about creating API tokens in Managed deployments, refer to the Dynatrace Managed documentation. Your API token must include the following scopes for full functionality:

• Access problem and event feed, metrics, and topology (DataExport)
• Read entities (entities.read)
• Read events (events.read)
• Read logs (logs.read)
• Read metrics (metrics.read)
• Read problems (problems.read)
• Read security problems (securityProblems.read)
• Read SLO (slo.read)

Configuration File

Configuration parameters

There are two ways to configure your Dynatrace Managed environments.

Method 1: Configuration File (Recommended for Local Development)

Example: dt-config.yaml

# Production environment
- apiEndpointUrl: https://my-api.company.com/
  environmentId: abc-123
  alias: production
  # Token is injected from an environment variable at runtime
  apiToken: ${DT_PROD_TOKEN}
  # You can also use the token directly
  # apiToken: dt0s01.ABCDEFGHIJK0123

# Staging environment
- apiEndpointUrl: https://staging-api.company.com/
  environmentId: xyz-789
  alias: staging
  apiToken: ${DT_STAGING_TOKEN}

Example: dt-config.json

[
  {
    "apiEndpointUrl": "https://my-api.company.com/",
    "environmentId": "abc-123",
    "alias": "production",
    "apiToken": "${DT_PROD_TOKEN}"
  }
]

Method 2: Environment Variable (Docker/Kubernetes)

For Kubernetes deployments or if you prefer environment variables, you can set DT_ENVIRONMENT_CONFIGS with a JSON string in either your .env file or directly in MCP Server connection configuration file

DT_ENVIRONMENT_CONFIGS='[{"apiEndpointUrl":"https://api.example.com/","environmentId":"abc-123","alias":"production","apiToken":"dt0s01.ABCDEFGHIJK0123"}]'

MCP Server connection configuration file

To actually connect to the MCP server you have to configure your MCP connection in your AI Assistant

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

VS Code

{
  "servers": {
    "npx-dynatrace-managed-mcp": {
      "command": "npx",
      "cwd": "${workspaceFolder}",
      "args": ["-y", "@dynatrace-oss/dynatrace-managed-mcp-server@latest"],
      "envFile": "${workspaceFolder}/.env"
    }
  }
}

Alternatively, this can also be stored in user settings, and you can define env as follows:

{
  "servers": {
    "npx-dynatrace-managed-mcp": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-managed-mcp-server@latest"],
      "env": {
        "DT_PROD_TOKEN": "dt0s01.ABCDEFGHIJK0123",
        "DT_CONFIG_FILE": "dt-config.yaml"
      }
    }
  }
}

Claude Code

Claude Code can install this server as a plugin, which skips the manual MCP configuration below and adds a skill covering environment selection and entity selectors. It is published through Anthropic's community plugin marketplace:

/plugin marketplace add anthropics/claude-plugins-community
/plugin install dynatrace-managed-mcp@claude-community

Claude Code prompts for the cluster configuration during install - either the JSON array for a single cluster, or a path to a dt-config.yaml / dt-config.json file for several. See docs/claude-code-plugin.md for details.

/plugin marketplace add dynatrace-oss/dynatrace-managed-mcp
/plugin install dynatrace-managed-mcp@dynatrace

To configure the server by hand instead, use the Claude Desktop snippet below - it applies to Claude Code's .mcp.json as well.

Claude Desktop

{
  "mcpServers": {
    "dynatrace-managed-mcp": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-managed-mcp-server@latest"],
      "env": {
        "DT_PROD_TOKEN": "dt0s01.ABCDEFGHIJK0123",
        "DT_CONFIG_FILE": "dt-config.yaml"
      }
    }
  }
}

Kiro

{
  "mcpServers": {
    "dynatrace-managed-mcp": {
      "command": "npx",
      "args": ["-y", "@dynatrace-oss/dynatrace-managed-mcp-server@latest"],
      "env": {
        "DT_PROD_TOKEN": "dt0s01.ABCDEFGHIJK0123",
        "DT_CONFIG_FILE": "dt-config.yaml"
      }
    }
  }
}

This configuration should be stored in <project-root>/.kiro/settings/mcp.json, or in user-level settings (~/.kiro/settings/mcp.json).

Google Gemini CLI

Using gemini CLI directly (recommended):

gemini extensions install https://github.com/dynatrace-oss/dynatrace-managed-mcp
export DT_ENVIRONMENT_CONFIGS="[{\"apiEndpointUrl\":\"https://my-api-endpoint.com/\",\"environmentId\":\"my-env-id-1\",\"alias\":\"alias-env\",\"apiToken\":\"my-api-token\"},{\"apiEndpointUrl\":\"https://my-api2-endpoint.com/\",\"environmentId\":\"my-env-id-2\",\"alias\":\"alias-env-2\",\"apiToken\":\"my-api-token-2\"}]"

and verify that the server is running via

gemini mcp list

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

{
  "mcpServers": {
    "dynatrace-managed-mcp": {
      "command": "npx",
      "args": ["@dynatrace-oss/dynatrace-managed-mcp-server@latest"],
      "env": {
        "DT_ENVIRONMENT_CONFIGS": "[{\"apiEndpointUrl\":\"https://my-api-endpoint.com/\",\"environmentId\":\"my-env-id-1\",\"alias\":\"alias-env\",\"apiToken\":\"my-api-token\"},{\"apiEndpointUrl\":\"https://my-api2-endpoint.com/\",\"environmentId\":\"my-env-id-2\",\"alias\":\"alias-env-2\",\"apiToken\":\"my-api-token-2\"}]",
        "DT_CONFIG_FILE": "dt-config.yaml"
      },
      "timeout": 30000,
      "trust": false
    }
  }
}

HTTP Server Mode (Alternative)

The default mode for this local MCP uses stdio for transport.

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

Running as HTTP server

Make sure you have the Configuration File in the same folder. You do not have to define API tokens for configurations ran in HTTP mode.

# Get help and see all available options
npx -y @dynatrace-oss/dynatrace-managed-mcp-server@latest --help

# Run with HTTP server on default port 3000
npx -y @dynatrace-oss/dynatrace-managed-mcp-server@latest --http

# Run with custom port
npx -y @dynatrace-oss/dynatrace-managed-mcp-server@latest --http --port 3001

# Run with custom host/IP
npx -y @dynatrace-oss/dynatrace-mcp-server@latest --http --host 127.0.0.1   # recommended for local computers
npx -y @dynatrace-oss/dynatrace-mcp-server@latest --http --host 0.0.0.0     # recommended for container
npx -y @dynatrace-oss/dynatrace-mcp-server@latest --http --host 192.168.0.1 # recommended when sharing connection over a local network

MCP Server connection configuration file:

As explained earlier, HTTP mode does not store API tokens in its configuration. Authentication is done by the user by filling the X-Dynatrace-Tokens header.

{
  "mcpServers": {
    "dynatrace-managed-mcp": {
      "url": "http://localhost:3000",
      "transport": "http",
      "headers": {
        "Content-Type": "application/json",
        "Accept": "application/json,text/event-stream",
        "X-Dynatrace-Tokens": "alias1=token1;alias2=token2"
      }
    }
  }
}

Performance Considerations

Important: This MCP server makes API calls to the Dynatrace Managed environment(s). It is designed for efficient usage (e.g., limiting the response sizes), but care should be taken not to overload the Dynatrace Managed environment(s) with large queries.

Best Practices:

1. Use specific time ranges (e.g., 1-2 hours) rather than large historical queries.
2. Use specific filters to limit the scope of queries as much as possible, for example, entity selectors that specify the entity ID.
3. If using multiple environments, be specific about which one to query, where applicable. If querying multiple at once, be mindful of how much data will be returned to the LLM, e.g. top 10 problems from 2 envs = 20 problems, versus top 10 problems from 10 envs = 100 problems.

DNS Rebinding Protection (HTTP mode)

• DT_MCP_ALLOWED_HOSTS (optional): Comma-separated list of hostnames the server accepts in the Host header. Ports are ignored, so list hostnames only (use the bracketed form for IPv6, e.g. [::1]).

When this variable is not set, the allowlist is derived from --host: the bound address plus localhost, 127.0.0.1 and [::1]. Requests whose Host header is not on the list are rejected with 403 Forbidden, as are requests carrying an Origin header for a hostname that is not on the list. This is what prevents DNS rebinding attacks.

Validation is always active: there is no configuration in which it is silently skipped.

Example: container bound to all interfaces, reached as mcp.internal.example.com:

DT_MCP_ALLOWED_HOSTS=mcp.internal.example.com node dist/index.js --http --host 0.0.0.0

DT_MCP_ALLOWED_HOSTS replaces the derived list rather than extending it, so include loopback names explicitly if you also need local access:

DT_MCP_ALLOWED_HOSTS=mcp.internal.example.com,localhost,127.0.0.1

Troubleshooting

Authentication Issues

In most cases, authentication issues stem from missing scopes or invalid tokens. Please ensure that you have added all required scopes as listed above.

When experiencing errors, you can ask the AI Assistant for the exact error returned by the MCP. For startup issues, check the AI Assistant logs.

You can also try running the MCP directly to see if it reports errors on startup:

```bash
npx @dynatrace-oss/dynatrace-managed-mcp-server@latest
```

Header size limits is too small

The X-Dynatrace-Tokens header grows with the number of environments. Each entry is roughly alias=dt0s01.ABCDEFGHIJK0123 (~110 characters). Node.js enforces a default HTTP header size limit of 16 KB, which accommodates approximately 140–150 environments before requests are rejected.

If you need more environments, increase the limit at server startup with the --max-http-header-size flag:

node --max-http-header-size=65536 ./dist/index.js --http

If you are running a reverse proxy (such as nginx) in front of the MCP server, the proxy also enforces its own limit. nginx defaults to 8 KB (large_client_header_buffers), which fits roughly 70 environments. Raise it in your nginx configuration:

large_client_header_buffers 4 32k;

Telemetry

The Dynatrace MCP Server includes sending Telemetry Data via Dynatrace OpenKit to help improve the product. This includes:

• Server start events
• Tool usage (which tools are called, success/failure, execution duration)
• Error tracking for debugging and improvement

Privacy and Opt-out:

• Telemetry is disabled by default but can be enabled by setting DT_MCP_ENABLE_TELEMETRY=true
• No sensitive data from your Dynatrace environment is tracked
• Only anonymous usage statistics and error information are collected
• Usage statistics and error data is transmitted to Dynatrace’s analytics endpoint

Configuration options:

• DT_MCP_ENABLE_TELEMETRY (boolean, default: false) - Enable Telemetry
• DT_MCP_TELEMETRY_APPLICATION_ID (string, default: dynatrace-managed-mcp) - Application ID for tracking
• DT_MCP_TELEMETRY_ENDPOINT_URL (string, default: Dynatrace endpoint) - OpenKit endpoint URL
• DT_MCP_TELEMETRY_DEVICE_ID (string, default: auto-generated) - Device identifier for tracking

Additional documentation

MCP Server usage

• API token scopes - table containing information about available tools, endpoints they are calling and required API token scopes to properly access them
• Architecture - detailed diagrams representing architecture of Dynatrace environment while using MCP server in either stdio or http mode
• Environment variables - detailed information about available environment variables
• Rule file - determine rules for your AI Assistant to ensure smooth usage of the Managed cluster

Development

• Changelog format - instruction for developers on how to write consistent and structured changelogs
• Development - general information about running the project and its contents