Don't worry if you can't use Datadog MCP! Use agent-skills × pup to investigate incidents using AI
This is Onkai Yuta (@fat47) from the Service Reliability Group (SRG) of the Media Headquarters.
#SRG(Service Reliability Group) is a group that mainly provides cross-sectional support for the infrastructure of our media services, improving existing services, launching new ones, and contributing to OSS.
This article summarizes the experience of running Datadog agent-skills and pup with Github Actions to automatically investigate issues.
I hope this helps in some way.
Datadog MCP Server continues in private previewDatadog Pup CLIInstalling Pup (for Mac)Try browser authentication with PupExample of operation with PupDatadog Skills for AI AgentsHow to install agent-skillsTry using agent-skills from Claude CodeLet's try to enable initial investigation using Datadog agent-skills × pup from Github ActionsOverall configuration diagramFirst, let me show you an example of the report output using Github Actions.Actual procedureConclusion
Datadog MCP Server continues in private preview
Datadog MCP Server was announced in private preview at DASH in June 2025.
I applied for the private preview of MCP Server immediately after the announcement, but before I knew it, the new year had come and I hadn't received an invitation.
Meanwhile,Datadog's official CLI tool, "Pup CLI," will be previewed in February.It was done.
Datadog Pup CLI
This is the official command line tool provided by Datadog, a comprehensive CLI that supports AI agents.
While the previous Datadog API required an API key, the Pup CLI supports OAuth2 authentication, allowing you to use browser-based authentication.
FORCE_AGENT_MODE=1Installing Pup (for Mac)
Try browser authentication with Pup
Select the organization you want to authenticate.
A list of permissions will be displayed, so accept them.
The authenticated token is valid for one hour.
If the token expires, you can refresh it with refresh.
Example of operation with Pup
The following operations are possible: (Partially quoted from the official website)
Monitors
Metrics
Dashboards
Datadog Skills for AI Agents
About a week after the initial release of Pup, Datadog agent-skills was released.
This is a "skill collection for teaching AI agents the steps to investigate with Datadog."
What to check and how to check using Pupis defined.
Specifically, the following skills are defined:
| Skill | Description |
|---|---|
| dd-pup | Defining authentication and commands in Pup |
| dd-monitors | Manage and mute monitors |
| dd-logs | Log Search |
| dd-apm | APM tracing etc. |
| dd-docs | Search the official Datadog documentation |
| dd-llmo | LLM Observability related (depending on Datadog MCP Server toolset) |
How to install agent-skills
Try using agent-skills from Claude Code
Let's try giving the following instructions in Claude Code.
APMでprd環境の◯◯サービスのパフォーマンスをチェック
After loading the dd-apm skill, you can see that the pup command is fetching the Datadog values.
In the end it displayed the following result:
Let's try to enable initial investigation using Datadog agent-skills × pup from Github Actions
I was able to confirm that I could use Claude Code to explore Datadog data from my own device.
Next, we tested whether we could use Github Actions to conduct an initial investigation.
Overall configuration diagram
The configuration is as follows: Install Datadog agent-skills and Pup from Github Actions, and use the Claude Code Action to run it using the Claude Sonnet 4.6 model on AWS Bedrock.
First, let me show you an example of the report output using Github Actions.
With this configuration, I manually executed Github Actions to output the report results.
First, an overall summary is displayed.
It summarizes slow endpoints and suggests specific actions to take for items that should be investigated immediately.
Actual procedure
Here we will introduce the steps to actually get it up and running.
In the case of Pup's OAuth2, the token expires in one hour, so this time we are using Pup by setting the Datadog API Key and APP_KEY.
Set the following environment variables in GitHub Secrets:
| Environment variable name | Value to set |
|---|---|
| Datadog API key | |
| Datadog APP key | |
| Datadog region (Japan),In the case of the US ) | |
| ARN of the IAM role for OIDC created in the AWS environment using Bedrock |
When issuing an APP key, it is possible to narrow down the operations that can be performed by using the Scope, so in this case we only granted read permissions for functions necessary for safety.
Create a yml file for the Github Actions workflow.
The model used this time is global.anthropic.claude-sonnet-4-6.
datadog-triage-claude.yml
Then run it manually from Actions.
In this example, we have made it possible to specify the APM service name and target period.
Running this will generate the report shown at the beginning of this chapter.
Conclusion
For a long time, it had been impossible to use Datadog MCP even if one wanted to, but the release of Pup CLI and agent-skill offered a ray of hope for the use of AI.
This time we tested it by manually running Github Actions, but it seems possible to use it in other applications such as integrating with Slack and launching it via a Webhook.
I'm going to improve it to make it even easier to use!
(I really want to use Datadog MCP soon!!!)
If you are interested in SRG, please contact us here.