Using skill at https://github.com/datadog-labs/agent-skills/blob/main/dd-orchestrator/SKILL.md setup Datadog RUM in my project

Overview

Agentic Onboarding is a set of AI-driven tools that automate Datadog instrumentation for your applications and infrastructure:

• AI Setup CLI: Set up Datadog from a terminal, without a coding assistant.
• MCP server: Set up Datadog through a coding assistant (such as Claude Code or Cursor), which handles framework detection and configuration from your IDE.

The two paths are complementary and use the same Datadog account. You can install the Datadog MCP Server in your IDE and run the CLI in a terminal.

AI Setup CLI

The Datadog AI Setup CLI is a standalone terminal tool. Use it when you don’t want to install an MCP server, or for tasks the MCP setup doesn’t support, such as creating a Datadog account.

The CLI can:

• Create a Datadog account end-to-end from the terminal
• Link an existing Datadog account to your local environment
• Instrument local infrastructure as code (Terraform, Helm, Kustomize, Ansible, Pulumi, raw Kubernetes manifests, Docker Compose files) by editing files in place
• Instrument local application code by adding SDK initialization and configuration for supported frontends and backends

Prerequisites

• Node.js 22 or later

Supported products

The CLI can set up the following products:

Install and run the CLI

Run the CLI with npx, passing --site to target your Datadog site (). You have two options, depending on whether you already have a Datadog account:

Configure and verify your setup

1. If the CLI can’t generate recommendations, or doesn’t find a strong match for your repository, it directs you to the full list of setup options. For direct setup with --product, you start at this menu.

2. The CLI detects your project’s frameworks, applies the required configuration, and provisions any necessary environment variables. Progress is reported stage by stage.

3. When setup completes, the CLI lists the products it instrumented and links to the Datadog UI to verify incoming data.

4. Commit the changes to your repository. You can edit the Datadog environment variables (API keys, application IDs) for your specific environment.

After the CLI completes, see the Next steps section to confirm data is flowing.

Headless mode

Headless mode is designed for unattended setups. An AI coding agent, CI job, or script can run the CLI directly in your repository and complete Datadog instrumentation on its own. No person needs to be present to approve prompts or make interactive choices.

Use --headless to skip the interactive UI. It requires both --site and --product:

DD_API_KEY=<API_KEY> DD_APP_KEY=<APP_KEY> \
  npx @datadog/ai-setup-cli \
  --headless \
  --site datadoghq.com \
  --product rum

Set the DD_API_KEY and DD_APP_KEY environment variables to authenticate without user interaction. Provide both variables together.

Alternatively, omit the API and application keys to authenticate with browser OAuth. OAuth is the only part of a headless run that might require user interaction, and it requires a localhost callback. For remote or fully unattended environments, use the DD_API_KEY and DD_APP_KEY environment variables instead.

By using --headless, you confirm that source-code upload and automatic command execution are authorized for the target project.

MCP Server

The Datadog MCP Server exposes the onboarding toolset to any MCP-compatible coding assistant. After you install and authenticate the server, you instrument a project by typing a one-line prompt. The agent reads your code, calls MCP tools (with your permission), applies changes, and verifies the result.

Prerequisites

• An MCP-compatible coding assistant, such as Claude Code or Cursor
• A Datadog account

Supported frameworks

Step 1: Install the MCP Server

claude mcp add --transport http datadog-onboarding- "?toolsets=onboarding"

{
  "mcpServers": {
    "datadog-onboarding-": {
      "url": "?toolsets=onboarding"
    }
  }
}

?toolsets=onboarding

Step 2: Authenticate the MCP Server

1. After you install the MCP Server, your coding assistant prompts you to authenticate. Press Enter to open the Datadog OAuth screen in your browser.
2. After authentication completes, choose Open to return to your IDE and grant the MCP Server access to your Datadog account.
3. Confirm that MCP tools appear under the datadog-onboarding- server.

Step 3: Instrument your project

Send the prompt that matches the product you want to set up:

Add Datadog Error Tracking to my project

Add Datadog Real User Monitoring to my project

Add Datadog Product Analytics to my project

Add Datadog for Kubernetes to my project

Add Datadog for Docker to my project

Add Datadog App and API Protection to my project

Add Datadog for AWS Lambda to my project

npx @datadog/ai-setup-cli --product serverless --serverless-compute-type=aws-lambda

Add Datadog for GCP Cloud Run containers to my project

npx @datadog/ai-setup-cli --product serverless --serverless-compute-type=gcp-cloud-run

Add Datadog for GCP Cloud Run functions to my project

npx @datadog/ai-setup-cli --product serverless --serverless-compute-type=gcp-cloud-run-functions

Add Datadog for Azure Container Apps to my project

npx @datadog/ai-setup-cli --product serverless --serverless-compute-type=azure-container-apps

The agent detects your stack, requests permission before each tool call, applies changes locally (without committing them), and prints verification steps.

After the agent completes, commit the changes to your repository and set any new environment variables (API keys, application IDs) in your production environment. Then see the Next steps section to confirm data is flowing.

Next steps

Confirm data is flowing in the Datadog UI for the product you set up:

• Error Tracking
• App and API Protection
• RUM > Applications
• Infrastructure > Hosts
• Serverless > Functions
• Logs > Live Tail

Further reading

Additional helpful documentation, links, and articles:

Instrument serverless apps with agentic onboardingBLOG