datadog:setup

Datadog Setup Skill

Interactive setup wizard for the Datadog plugin. Verifies API credentials, tests MCP connectivity, and configures project-level defaults for log queries and dashboards.

Tools Used

• Read: Check ~/.claude/settings.local.json for environment variables
• ToolSearch: Discover Datadog MCP tools and verify connectivity
• AskUserQuestion: Prompt for missing configuration and preferences
• Edit: Update project's CLAUDE.md with default settings

Workflow

Step 1: Check Environment Variables

Read ~/.claude/settings.local.json and verify the presence of required Datadog credentials in the env object:

Variable	Required	Description
DD_API_KEY	Yes	Datadog API key for authentication
DD_APP_KEY	Yes	Datadog Application key for API access
DD_SITE	Yes	Datadog site region endpoint

For each missing variable:

1. Explain what it is and where to find it:

   • DD_API_KEY: Datadog → Organization Settings → API Keys
   • DD_APP_KEY: Datadog → Organization Settings → Application Keys (needs logs_read_data, dashboards_read, metrics_read scopes)
   • DD_SITE: The Datadog site for your organization

2. Show common DD_SITE values:

   Site	Value
   EU	datadoghq.eu
   US1	datadoghq.com
   US5	us5.datadoghq.com
   US3	us3.datadoghq.com
   AP1	ap1.datadoghq.com

IMPORTANT: Do NOT write credentials automatically. Only instruct the user on what to add and where. The user must edit the file themselves.

3. Show the exact JSON snippet to add to ~/.claude/settings.local.json:

   {
     "env": {
       "DD_API_KEY": "your-api-key-here",
       "DD_APP_KEY": "your-app-key-here",
       "DD_SITE": "datadoghq.eu"
     }
   }
   

If all variables are present, confirm with a checkmark and proceed to Step 2.

Step 2: Verify Connection

Test that the Datadog MCP server is reachable and credentials are valid:

1. Use ToolSearch("datadog") to discover available MCP tools
2. Attempt a lightweight API call (e.g., list-metrics with a small limit or get-metrics with a known metric prefix) to verify credentials work
3. If successful: Report connected site and available tools
4. If failed: Diagnose the issue:
   • No MCP tools found → Plugin not installed correctly, suggest reinstalling
   • Authentication error → Wrong API key, expired key, or insufficient scopes
   • Connection error → Wrong DD_SITE value, network issues
   • Suggest specific fixes for each failure mode

Step 3: Configure Default Service (Optional)

Ask the user if they want to configure a default Datadog service for this project:

"Do you have a default Datadog service name for this project? This will be used to auto-scope log queries."

• If yes: Get the service name and write it to the project's CLAUDE.md
• If no/skip: Proceed without a default (queries will require explicit service specification)

The service name should match what appears in Datadog's Service Catalog (e.g., backend-api, frontend-app, worker).

Step 4: Configure Default Search Filter (Optional)

Ask the user if they want a default search filter prepended to all log queries:

"Do you want a default filter prepended to all log queries? Common examples: env:production, team:backend, kube_namespace:production"

• If yes: Get the filter expression and write it to the project's CLAUDE.md
• If no/skip: Proceed without a default filter

Step 5: Write Configuration to CLAUDE.md

If the user configured any defaults (service or filter), append the configuration block to the project's local CLAUDE.md file in the current working directory.

Format (using HTML comments so it's invisible in rendered markdown but parseable):

<!-- datadog-plugin-config -->
<!-- datadog_default_service: my-service-name -->
<!-- datadog_default_filter: env:production -->

Rules:

• If CLAUDE.md exists, append the config block at the end
• If CLAUDE.md does not exist, create it with only the config block
• If a <!-- datadog-plugin-config --> block already exists, replace it
• Only include lines for settings the user actually configured

Note: This configuration format is read by datadog:logs (Step 1: Detect Context) to auto-scope queries. Changes to the format must be synchronized between both skills.

Step 6: Summary

Print a configuration summary:

Datadog Plugin Configuration
-----------------------------
DD_SITE:          datadoghq.eu (configured)
DD_API_KEY:       configured
DD_APP_KEY:       configured
MCP Connection:   verified
Default service:  backend-api
Default filter:   env:production

Next steps:
- Search logs: /datadog:logs show me recent errors
- Analyze traces: /datadog:traces find recent error traces
- Create dashboard: /datadog:dashboard create a Redis dashboard

Adjust the summary based on what was actually configured. For missing items, show "not configured" and suggest running setup again.

Error Handling

Situation	Action
~/.claude/settings.local.json doesn't exist	Tell user to create it with the required env vars
JSON parse error in settings file	Warn user about malformed JSON, suggest fixing
MCP server not responding	Check if plugin is installed, suggest /plugin install datadog@fabn-claude-plugins
API key has insufficient scopes	List required scopes: logs_read_data, dashboards_read, metrics_read
CLAUDE.md is read-only	Warn and print the config block for manual addition

Related Skills

• /datadog:logs — Search and analyze logs (reads configuration written by this skill)
• /datadog:traces — Search and analyze APM traces (reads configuration written by this skill)
• /datadog:dashboard — Create Terraform-based Datadog dashboards